| {: , : {: {: [{: , : 84, : 88}, {: , : 92, : 101}, {: , : 148, : 157}, {: , : 167, : 189}, {: , : 193, : 200}, {: , : 210, : 220}, {: , : 235, : 244}, {: , : 245, : 254}, {: , : 273, : 287}, {: , : 305, : 319}, {: , : 329, : 340}, {: , : 355, : 364}, {: , : 365, : 375}, {: , : 408, : 416}, {: , : 438, : 442}, {: , : 446, : 455}, {: , : 473, : 478}, {: , : 503, : 516}, {: , : 543, : 548}, {: , : 549, : 559}, {: , : 568, : 574}, {: , : 575, : 581}, {: , : 597, : 603}, {: , : 615, : 624}, {: , : 645, : 649}, {: , : 657, : 667}, {: , : 675, : 683}, {: , : 739, : 750}, {: , : 756, : 761}, {: , : 762, : 773}, {: , : 792, : 810}, {: , : 821, : 831}, {: , : 884, : 904}, {: , : 940, : 947}, {: , : 951, : 962}, {: , : 982, : 994}, {: , : 998, : 1006}, {: , : 1017, : 1026}, {: , : 1056, : 1070}, {: , : 1074, : 1083}, {: , : 1088, : 1096}, {: , : 1101, : 1108}, {: , : 1114, : 1123}, {: , : 1147, : 1163}, {: , : 1164, : 1175}]}, : {: [{: {: , : 235, : 244}, : {: , : 210, : 220}}, {: {: , : 1056, : 1070}, : {: , : 1074, : 1083}}]}}, : []} |
| {: , : {: {: [{: , : 64, : 69}, {: , : 107, : 116}, {: , : 136, : 141}, {: , : 160, : 166}, {: , : 206, : 216}, {: , : 246, : 254}, {: , : 297, : 302}, {: , : 322, : 329}, {: , : 374, : 378}, {: , : 430, : 434}, {: , : 461, : 468}, {: , : 524, : 535}, {: , : 541, : 549}, {: , : 570, : 575}, {: , : 595, : 602}, {: , : 626, : 637}, {: , : 648, : 655}, {: , : 665, : 673}, {: , : 702, : 713}]}, : {: [{: {: , : 64, : 69}, : {: , : 136, : 141}}, {: {: , : 595, : 602}, : {: , : 541, : 549}}]}}, : []} |
| {: pluggable \, : {: {: [{: , : 75, : 80}, {: , : 81, : 86}, {: , : 96, : 103}, {: , : 107, : 116}, {: , : 154, : 167}, {: , : 172, : 176}, {: , : 177, : 186}, {: , : 187, : 203}, {: , : 242, : 251}, {: , : 256, : 268}, {: , : 282, : 289}, {: , : 294, : 299}, {: , : 313, : 325}, {: , : 330, : 341}, {: , : 349, : 359}, {: , : 383, : 392}, {: , : 393, : 404}, {: , : 421, : 430}, {: , : 439, : 449}, {: , : 468, : 473}, {: , : 486, : 493}, {: , : 498, : 506}, {: , : 542, : 546}, {: , : 547, : 557}, {: , : 559, : 568}, {: , : 614, : 622}, {: , : 623, : 632}, {: , : 692, : 697}, {: , : 707, : 716}]}, : {: [{: {: , : 154, : 167}, : {: , : 187, : 203}}]}}, : []} |
| {: , : {: {: [{: , : 57, : 65}, {: , : 76, : 89}, {: , : 93, : 112}, {: , : 131, : 136}, {: , : 153, : 164}, {: , : 173, : 186}, {: , : 194, : 209}, {: , : 210, : 214}, {: , : 219, : 228}, {: , : 235, : 246}, {: , : 273, : 282}, {: , : 287, : 295}, {: , : 329, : 337}, {: , : 353, : 362}, {: , : 393, : 399}, {: , : 403, : 412}, {: , : 417, : 422}, {: , : 440, : 466}, {: , : 467, : 478}, {: , : 483, : 492}, {: , : 519, : 531}, {: , : 536, : 549}, {: , : 555, : 565}, {: , : 566, : 576}, {: , : 629, : 639}, {: , : 678, : 690}, {: , : 707, : 726}, {: , : 773, : 778}, {: , : 790, : 798}, {: , : 846, : 857}, {: , : 867, : 879}, {: , : 904, : 912}, {: , : 916, : 926}, {: , : 927, : 934}, {: , : 955, : 964}, {: , : 1004, : 1011}, {: , : 1025, : 1030}, {: , : 1036, : 1048}, {: , : 1049, : 1056}, {: , : 1097, : 1116}, {: , : 1130, : 1140}, {: , : 1151, : 1158}, {: , : 1181, : 1199}, {: , : 1216, : 1222}]}, : {: [{: {: , : 153, : 164}, : {: , : 173, : 186}}, {: {: , : 467, : 478}, : {: , : 519, : 531}}]}}, : []} |
| {: , : {: {: [{: , : 123, : 128}, {: , : 161, : 169}, {: , : 173, : 178}, {: , : 194, : 202}, {: , : 221, : 227}, {: , : 228, : 239}, {: , : 252, : 260}, {: , : 288, : 294}, {: , : 319, : 326}, {: , : 360, : 368}, {: , : 392, : 399}, {: , : 406, : 413}, {: , : 414, : 420}, {: , : 421, : 428}, {: , : 468, : 476}, {: , : 477, : 488}, {: , : 519, : 525}, {: , : 526, : 535}, {: , : 556, : 564}, {: , : 581, : 586}, {: , : 605, : 612}, {: , : 621, : 631}, {: , : 644, : 652}, {: , : 653, : 659}, {: , : 678, : 687}, {: , : 706, : 715}, {: , : 736, : 747}, {: , : 748, : 756}, {: , : 781, : 789}, {: , : 807, : 810}, {: , : 817, : 835}, {: , : 856, : 862}, {: , : 890, : 900}, {: , : 902, : 930}, {: , : 935, : 946}]}, : {: [{: {: , : 194, : 202}, : {: , : 221, : 227}}, {: {: , : 526, : 535}, : {: , : 477, : 488}}, {: {: , : 581, : 586}, : {: , : 605, : 612}}]}}, : []} |
| {: , : {: {: [{: , : 129, : 140}, {: , : 142, : 153}, {: , : 171, : 183}, {: , : 222, : 226}, {: , : 230, : 241}, {: , : 242, : 253}, {: , : 260, : 267}, {: , : 301, : 307}, {: , : 308, : 317}, {: , : 377, : 387}, {: , : 412, : 424}, {: , : 470, : 479}, {: , : 514, : 520}, {: , : 521, : 530}, {: , : 534, : 545}, {: , : 631, : 640}, {: , : 665, : 677}, {: , : 702, : 713}, {: , : 715, : 724}, {: , : 729, : 736}, {: , : 743, : 749}, {: , : 795, : 807}, {: , : 810, : 816}, {: , : 832, : 840}, {: , : 844, : 856}, {: , : 881, : 893}, {: , : 908, : 919}, {: , : 926, : 931}, {: , : 932, : 939}, {: , : 947, : 955}, {: , : 959, : 963}]}, : {: [{: {: , : 832, : 840}, : {: , : 844, : 856}}]}}, : []} |
| {: , : {: {: [{: , : 132, : 139}, {: , : 211, : 218}, {: , : 236, : 250}, {: , : 254, : 260}, {: , : 295, : 322}, {: , : 351, : 362}, {: , : 366, : 371}, {: , : 372, : 378}, {: , : 383, : 391}, {: , : 398, : 403}, {: , : 404, : 411}, {: , : 419, : 425}, {: , : 434, : 441}, {: , : 448, : 453}, {: , : 454, : 468}, {: , : 469, : 479}, {: , : 490, : 496}, {: , : 502, : 508}, {: , : 509, : 516}, {: , : 531, : 537}, {: , : 564, : 573}, {: , : 617, : 622}, {: , : 634, : 639}, {: , : 640, : 654}, {: , : 659, : 667}, {: , : 686, : 692}, {: , : 707, : 732}, {: , : 739, : 745}, {: , : 757, : 763}, {: , : 768, : 776}, {: , : 781, : 795}, {: , : 796, : 804}, {: , : 810, : 816}, {: , : 825, : 835}, {: , : 861, : 865}, {: , : 891, : 897}, {: , : 915, : 923}, {: , : 939, : 945}, {: , : 965, : 970}, {: , : 971, : 985}, {: , : 1002, : 1010}, {: , : 1047, : 1061}, {: , : 1062, : 1070}, {: , : 1099, : 1113}, {: , : 1114, : 1121}, {: , : 1160, : 1165}, {: , : 1166, : 1178}]}, : {: [{: {: , : 236, : 250}, : {: , : 254, : 260}}, {: {: , : 490, : 496}, : {: , : 469, : 479}}, {: {: , : 564, : 573}, : {: , : 502, : 508}}, {: {: , : 707, : 732}, : {: , : 739, : 745}}, {: {: , : 939, : 945}, : {: , : 1002, : 1010}}]}}, : []} |
| {: , : {: {: [{: , : 70, : 85}, {: , : 109, : 117}, {: , : 121, : 148}, {: , : 156, : 169}, {: , : 229, : 234}, {: , : 273, : 292}, {: , : 297, : 305}, {: , : 315, : 320}, {: , : 329, : 351}, {: , : 356, : 364}, {: , : 382, : 388}, {: , : 405, : 410}, {: , : 411, : 420}, {: , : 421, : 431}, {: , : 436, : 444}, {: , : 445, : 451}, {: , : 457, : 464}, {: , : 479, : 489}, {: , : 507, : 519}, {: , : 539, : 550}, {: , : 561, : 571}, {: , : 587, : 594}, {: , : 624, : 632}, {: , : 633, : 640}, {: , : 641, : 648}]}, : {: [{: {: , : 587, : 594}, : {: , : 561, : 571}}]}}, : []} |
| {: , : {: {: [{: , : 68, : 76}, {: , : 89, : 94}, {: , : 95, : 101}, {: , : 102, : 113}, {: , : 135, : 142}, {: , : 157, : 179}, {: , : 180, : 183}, {: , : 202, : 211}, {: , : 215, : 220}, {: , : 228, : 235}, {: , : 250, : 257}, {: , : 258, : 268}, {: , : 275, : 280}, {: , : 292, : 301}, {: , : 331, : 335}, {: , : 336, : 344}, {: , : 345, : 354}, {: , : 359, : 367}, {: , : 394, : 401}, {: , : 429, : 434}, {: , : 444, : 451}, {: , : 460, : 465}, {: , : 490, : 496}, {: , : 498, : 523}, {: , : 528, : 543}, {: , : 553, : 558}, {: , : 570, : 577}, {: , : 586, : 593}, {: , : 603, : 608}, {: , : 609, : 614}, {: , : 630, : 640}, {: , : 656, : 667}, {: , : 689, : 695}, {: , : 697, : 708}, {: , : 730, : 741}, {: , : 787, : 791}, {: , : 809, : 816}]}, : {: [{: {: , : 202, : 211}, : {: , : 215, : 220}}, {: {: , : 697, : 708}, : {: , : 730, : 741}}]}}, : []} |
| {: , : {: {: [{: , : 83, : 91}, {: , : 107, : 112}, {: , : 127, : 133}, {: , : 134, : 147}, {: , : 157, : 164}, {: , : 165, : 173}, {: , : 185, : 193}, {: , : 210, : 216}, {: , : 223, : 230}, {: , : 234, : 240}, {: , : 241, : 259}, {: , : 269, : 277}, {: , : 297, : 311}, {: , : 329, : 335}, {: , : 336, : 347}, {: , : 361, : 369}, {: , : 370, : 376}, {: , : 377, : 383}, {: , : 430, : 437}, {: , : 459, : 473}, {: , : 478, : 484}, {: , : 485, : 491}, {: , : 492, : 514}, {: , : 518, : 534}, {: , : 571, : 574}, {: , : 588, : 613}, {: , : 614, : 624}, {: , : 625, : 629}, {: , : 656, : 664}, {: , : 677, : 687}, {: , : 701, : 711}, {: , : 728, : 749}, {: , : 785, : 808}, {: , : 820, : 828}, {: , : 865, : 878}, {: , : 885, : 892}, {: , : 923, : 930}, {: , : 971, : 984}, {: , : 997, : 1005}, {: , : 1045, : 1049}, {: , : 1050, : 1056}]}, : {: [{: {: , : 241, : 259}, : {: , : 83, : 91}}, {: {: , : 336, : 347}, : {: , : 269, : 277}}, {: {: , : 430, : 437}, : {: , : 361, : 369}}, {: {: , : 865, : 878}, : {: , : 923, : 930}}]}}, : []} |
| {: , : {: {: [{: , : 61, : 67}, {: , : 72, : 76}, {: , : 77, : 84}, {: , : 104, : 115}, {: , : 138, : 163}, {: , : 170, : 175}, {: , : 176, : 183}, {: , : 191, : 203}, {: , : 208, : 218}, {: , : 237, : 251}, {: , : 252, : 258}, {: , : 263, : 267}, {: , : 268, : 275}, {: , : 277, : 291}, {: , : 296, : 307}, {: , : 320, : 333}, {: , : 334, : 341}, {: , : 363, : 368}, {: , : 374, : 387}, {: , : 415, : 425}, {: , : 415, : 425}, {: , : 436, : 440}, {: , : 441, : 447}, {: , : 499, : 513}, {: , : 519, : 534}, {: , : 542, : 548}, {: , : 580, : 585}, {: , : 591, : 604}, {: , : 632, : 642}, {: , : 643, : 648}, {: , : 683, : 693}, {: , : 698, : 706}, {: , : 712, : 737}, {: , : 742, : 747}, {: , : 751, : 755}, {: , : 766, : 779}, {: , : 797, : 807}, {: , : 809, : 823}, {: , : 824, : 835}, {: , : 841, : 850}, {: , : 854, : 861}, {: , : 866, : 872}, {: , : 873, : 884}, {: , : 895, : 901}, {: , : 902, : 909}, {: , : 912, : 927}, {: , : 963, : 969}]}, : {: [{: {: , : 77, : 84}, : {: , : 138, : 163}}, {: {: , : 363, : 368}, : {: , : 334, : 341}}, {: {: , : 415, : 425}, : {: , : 374, : 387}}, {: {: , : 499, : 513}, : {: , : 441, : 447}}, {: {: , : 712, : 737}, : {: , : 751, : 755}}, {: {: , : 895, : 901}, : {: , : 912, : 927}}]}}, : []} |
| {: , : {: {: [{: , : 78, : 83}, {: , : 84, : 91}, {: , : 128, : 136}, {: , : 137, : 155}, {: , : 159, : 165}, {: , : 192, : 203}, {: , : 214, : 223}, {: , : 224, : 232}, {: , : 244, : 254}, {: , : 267, : 277}, {: , : 305, : 320}, {: , : 331, : 339}, {: , : 352, : 358}, {: , : 366, : 370}, {: , : 374, : 383}, {: , : 384, : 398}, {: , : 402, : 420}, {: , : 426, : 435}, {: , : 440, : 447}, {: , : 455, : 460}, {: , : 466, : 476}, {: , : 485, : 489}, {: , : 553, : 563}, {: , : 586, : 595}, {: , : 596, : 604}, {: , : 626, : 633}, {: , : 651, : 668}, {: , : 673, : 686}, {: , : 692, : 703}]}, : {: [{: {: , : 78, : 83}, : {: , : 137, : 155}}, {: {: , : 352, : 358}, : {: , : 384, : 398}}]}}, : []} |
| {: Semantic annotation of text requires the dynamic merging of linguistically structured information and a \, usually represented as a domain-specific ontology. On the other hand, the process of engineering a domain ontology through semi-automatic ontology learning system requires the availability of a considerable amount of semantically annotated documents. Facing this bootstrapping paradox requires an incremental process of annotation-acquisition-annotation, whereby domain-specific knowledge is acquired from linguistically-annotated texts and then projected back onto texts for extra linguistic information to be annotated and further knowledge layers to be extracted. The presented methodology is a first step in the direction of a full \ circle where the semantic annotation platform and the evolving ontology interact in symbiosis. As a case study we have chosen the semantic annotation of product catalogues. We propose a hybrid approach, combining pattern matching techniques to exploit the regular structure of product descriptions in catalogues, and Natural Language Processing techniques which are resorted to analyze natural language descriptions. The semantic annotation involves the access to the ontology, semi-automatically bootstrapped with an ontology learning tool from annotated collections of catalogues.outputentitieskeyphrasetextSemanticstartendtexttextstartendtextrequiresstartendtextstructuredstartendtextinformationstartendtextmodelstartendtextdomain-specificstartendtextontologystartendtexthandstartendtextprocessstartendtextdomainstartendtextontologystartendtextsemi-automaticstartendtextontologystartendtextsystemstartendtextrequiresstartendtextavailabilitystartendtextamountstartendtextdocumentsstartendtextbootstrappingstartendtextrequiresstartendtextprocessstartendtextannotation-acquisition-annotationstartendtextdomain-specificstartendtextknowledgestartendtexttextsstartendtextprojectedstartendtexttextsstartendtextlinguistic informationstartendtextknowledgestartendtextlayersstartendtextextractedstartendtextmethodologystartendtextfirst stepstartendtextdirectionstartendtextsemanticstartendtextplatformstartendtextontologystartendtextcase studystartendtextsemanticstartendtextproductstartendtextproposestartendtextapproachstartendtextpatternstartendtextmatchingstartendtexttechniquesstartendtextstructurestartendtextproductstartendtextdescriptionsstartendtextNatural Language Processing techniquesstartendtextnatural languagestartendtextdescriptionsstartendtextsemanticstartendtextaccessstartendtextontologystartendtextontologystartendtexttoolstartendtextcollectionsstartendrelationsno_relationheadtextdocumentsstartendtailtextsystemstartendheadtextknowledgestartendtailtexttextsstartendheadtextstructurestartendtailtexttechniquesstartendheadtextdescriptionsstartendtailtextNatural Language Processing techniquesstartendheadtextcollectionsstartendtailtexttoolstartendschema |
| inputAnaphoric Annotation in the ARRAU Corpus Arrau is a new corpus annotated for anaphoric relations, with information about agreement and explicit representation of multiple antecedents for ambiguous anaphoric expressions and discourse antecedents for expressions which refer to abstract entities such as events, actions and plans. The corpus contains texts from different genres: task-oriented dialogues from the Trains-91 and Trains-93 corpus, narratives from the English Pear Stories corpus, newspaper articles from the Wall Street Journal portion of the Penn Treebank, and mixed text from the Gnome corpus.outputentitieskeyphrasetextcorpusstartendtextrelationsstartendtextinformationstartendtextagreementstartendtextrepresentationstartendtextexpressionsstartendtextdiscoursestartendtextexpressionsstartendtextabstractstartendtextentitiesstartendtexteventsstartendtextactionsstartendtextcorpusstartendtexttextsstartendtextgenresstartendtexttask-orientedstartendtextdialoguesstartendtextcorpusstartendtextEnglishstartendtextPearstartendtextcorpusstartendtextnewspaperstartendtextWall Street JournalstartendtextportionstartendtextPenn Treebankstartendtexttextstartendtextcorpusstartendrelationsno_relationheadtextcorpusstartendtailtextrelationsstartendheadtexttextsstartendtailtextcorpusstartendschema |
| inputImproving Probabilistic Latent Semantic Analysis With Principal Component Analysis Probabilistic Latent Semantic Analysis (PLSA) models have been shown to provide a better model for capturing polysemy and synonymy than Latent Semantic Analysis (LSA). However, the parameters of a PLSA model are trained using the Expectation Maximization (EM) algorithm, and as a result, the trained model is dependent on the initialization values so that performance can be highly variable. In this paper we present a method for using LSA analysis to initialize a PLSA model. We also investigated the performance of our method for the tasks of text segmentation and retrieval on personal-size corpora, and present results demonstrating the efficacy of our proposed approach.outputentitieskeyphrasetextProbabilistic Latent Semantic AnalysisstartendtextmodelsstartendtextprovidestartendtextmodelstartendtextLatent Semantic AnalysisstartendtextparametersstartendtextmodelstartendtexttrainedstartendtextExpectationstartendtextMaximizationstartendtextalgorithmstartendtextresultstartendtexttrainedstartendtextmodelstartendtextperformancestartendtextvariablestartendtextpaperstartendtextmethodstartendtextanalysisstartendtextmodelstartendtextperformancestartendtextmethodstartendtexttasksstartendtexttextstartendtextretrievalstartendtextpersonal-sizestartendtextcorporastartendtextresultsstartendtextproposedstartendtextapproachstartendrelationsno_relationheadtextalgorithmstartendtailtextmodelstartendheadtextpaperstartendtailtextmethodstartendschema |
| inputImproved Lexical Alignment By Combining Multiple Reified Alignments We describe a word alignment platform which ensures text pre-processing (to-kenization, POS-tagging, lemmatization, chunking, sentence alignment) as required by an accurate word alignment. The platform combines two different methods, producing distinct alignments. The basic word aligners are described in some details and are individually evaluated. The union of the individual alignments is subject to a filtering postprocessing phase. Two different filtering methods are also presented. The evaluation shows that the combined word alignment contains 10. 75% less errors than the best individual aligner.outputentitieskeyphrasetextword alignmentstartendtextplatformstartendtexttextstartendtextpre-processingstartendtextPOS-taggingstartendtextchunkingstartendtextsentencestartendtextalignmentstartendtextrequiredstartendtextword alignmentstartendtextplatformstartendtextmethodsstartendtextalignmentsstartendtextbasicstartendtextwordstartendtextdetailsstartendtextevaluatedstartendtextindividualstartendtextalignmentsstartendtextphasestartendtextmethodsstartendtextevaluationstartendtextword alignmentstartendtexterrorsstartendtextindividualstartendrelationsno_relationheadtextplatformstartendtailtextpre-processingstartendheadtextword alignmentstartendtailtexterrorsstartendschema |
| inputToward A Task-Based Gold Standard For Evaluation Of NP Chunks And Technical Terms We propose a gold standard for evaluating two types of information extraction output--noun phrase (NP) chunks (Abney 1991; Ramshaw and Marcus 1995) and technical terms (Justeson and Katz 1995; Daille 2000; Jacquemin 2002). The gold standard is built around the notion that since different semantic and syntactic variants of terms are arguably correct, a fully satisfactory assessment of the quality of the output must include task-based evaluation. We conducted an experiment that assessed subjects' choice of index terms in an information access task. Subjects showed significant preference for index terms that are longer, as measured by number of words, and more complex, as measured by number of prepositions. These terms, which were identified by a human indexer, serve as the gold standard. The experimental protocol is a reliable and rigorous method for evaluating the quality of a set of terms. An important advantage of this task-based evaluation is that a set of index terms which is different than the gold standard can' win' by providing better information access than the gold standard itself does. And although the individual human subject experiments are time consuming, the experimental interface, test materials anddata analysis programs are completely re-usable.", "output": {"entities": {"keyphrase": [{"text": "propose", "start": 85, "end": 92}, {"text": "gold standard", "start": 95, "end": 108}, {"text": "evaluating", "start": 113, "end": 123}, {"text": "types", "start": 128, "end": 133}, {"text": "information extraction", "start": 137, "end": 159}, {"text": "output", "start": 160, "end": 166}, {"text": "noun phrase", "start": 168, "end": 179}, {"text": "chunks", "start": 185, "end": 191}, {"text": "terms", "start": 244, "end": 249}, {"text": "gold standard", "start": 309, "end": 322}, {"text": "notion", "start": 343, "end": 349}, {"text": "semantic", "start": 371, "end": 379}, {"text": "syntactic", "start": 384, "end": 393}, {"text": "variants", "start": 394, "end": 402}, {"text": "terms", "start": 406, "end": 411}, {"text": "assessment", "start": 455, "end": 465}, {"text": "quality", "start": 473, "end": 480}, {"text": "output", "start": 488, "end": 494}, {"text": "include", "start": 500, "end": 507}, {"text": "task-based", "start": 508, "end": 518}, {"text": "evaluation", "start": 519, "end": 529}, {"text": "experiment", "start": 547, "end": 557}, {"text": "choice", "start": 582, "end": 588}, {"text": "index", "start": 592, "end": 597}, {"text": "terms", "start": 598, "end": 603}, {"text": "information access", "start": 610, "end": 628}, {"text": "task", "start": 629, "end": 633}, {"text": "preference", "start": 663, "end": 673}, {"text": "index", "start": 678, "end": 683}, {"text": "terms", "start": 684, "end": 689}, {"text": "number", "start": 722, "end": 728}, {"text": "words", "start": 732, "end": 737}, {"text": "complex", "start": 748, "end": 755}, {"text": "number", "start": 772, "end": 778}, {"text": "prepositions", "start": 782, "end": 794}, {"text": "terms", "start": 802, "end": 807}, {"text": "gold standard", "start": 864, "end": 877}, {"text": "experimental", "start": 883, "end": 895}, {"text": "protocol", "start": 896, "end": 904}, {"text": "method", "start": 932, "end": 938}, {"text": "evaluating", "start": 943, "end": 953}, {"text": "quality", "start": 958, "end": 965}, {"text": "terms", "start": 978, "end": 983}, {"text": "advantage", "start": 998, "end": 1007}, {"text": "task-based", "start": 1016, "end": 1026}, {"text": "evaluation", "start": 1027, "end": 1037}, {"text": "index", "start": 1055, "end": 1060}, {"text": "terms", "start": 1061, "end": 1066}, {"text": "gold standard", "start": 1095, "end": 1108}, {"text": "providing", "start": 1122, "end": 1131}, {"text": "information access", "start": 1139, "end": 1157}, {"text": "gold standard", "start": 1167, "end": 1180}, {"text": "individual", "start": 1211, "end": 1221}, {"text": "experiments", "start": 1236, "end": 1247}, {"text": "time", "start": 1252, "end": 1256}, {"text": "experimental", "start": 1272, "end": 1284}, {"text": "interface", "start": 1285, "end": 1294}, {"text": "test", "start": 1296, "end": 1300}, {"text": "anddata", "start": 1311, "end": 1318}, {"text": "analysis", "start": 1319, "end": 1327}, {"text": "programs", "start": 1328, "end": 1336}]}, "relations": {"no_relation": [{"head": {"text": "gold standard", "start": 95, "end": 108}, "tail": {"text": "information extraction", "start": 137, "end": 159}}, {"head": {"text": "variants", "start": 394, "end": 402}, "tail": {"text": "terms", "start": 406, "end": 411}}, {"head": {"text": "terms", "start": 802, "end": 807}, "tail": {"text": "gold standard", "start": 864, "end": 877}}, {"head": {"text": "method", "start": 932, "end": 938}, "tail": {"text": "evaluating", "start": 943, "end": 953}}]}}, "schema": []} |
| {"input": "Acquiring Inference Rules With Temporal Constraints By Using Japanese Coordinated Sentences And Noun-Verb Co-Occurrences \" This paper shows that inference rules with temporal constraints can be acquired by using verb-verb co-occurrences in Japanese coordinated sentences and verb-noun cooccurrences. For example, our unsuper-vised acquisition method could obtain the inference rule \" If someone enforces a law, usually someone enacts the law at the same time as or before the enforcing of the law \" since the verbs \" enact \" and \" enforce \" frequently co-occurred in coordinated sentences and the verbs also frequently co-occurred with the noun \" law \". We also show that the accuracy of the acquisition is improved by using the occurrence frequency of a single verb, which we assume indicates how generic the meaning of the verb is.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 128, "end": 133}, {"text": "inference", "start": 145, "end": 154}, {"text": "rules", "start": 155, "end": 160}, {"text": "constraints", "start": 175, "end": 186}, {"text": "verb-verb", "start": 212, "end": 221}, {"text": "Japanese", "start": 240, "end": 248}, {"text": "sentences", "start": 261, "end": 270}, {"text": "verb-noun", "start": 275, "end": 284}, {"text": "example", "start": 304, "end": 311}, {"text": "acquisition", "start": 331, "end": 342}, {"text": "method", "start": 343, "end": 349}, {"text": "inference", "start": 367, "end": 376}, {"text": "rule", "start": 377, "end": 381}, {"text": "time", "start": 454, "end": 458}, {"text": "verbs", "start": 509, "end": 514}, {"text": "sentences", "start": 579, "end": 588}, {"text": "verbs", "start": 597, "end": 602}, {"text": "noun", "start": 640, "end": 644}, {"text": "accuracy", "start": 676, "end": 684}, {"text": "acquisition", "start": 692, "end": 703}, {"text": "improved", "start": 707, "end": 715}, {"text": "occurrence", "start": 729, "end": 739}, {"text": "frequency", "start": 740, "end": 749}, {"text": "verb", "start": 762, "end": 766}, {"text": "meaning", "start": 810, "end": 817}, {"text": "verb", "start": 825, "end": 829}]}, "relations": {"no_relation": [{"head": {"text": "verbs", "start": 509, "end": 514}, "tail": {"text": "sentences", "start": 579, "end": 588}}, {"head": {"text": "frequency", "start": 740, "end": 749}, "tail": {"text": "verb", "start": 762, "end": 766}}]}}, "schema": []} |
| {"input": "Improved Affinity Graph Based Multi-Document Summarization This paper describes an affinity graph based approach to multi-document summarization. We incorporate a diffusion process to acquire semantic relationships between sentences, and then compute information richness of sentences by a graph rank algorithm on differentiated in-tra-document links and inter-document links between sentences. A greedy algorithm is employed to impose diversity penalty on sentences and the sentences with both high information richness and high information novelty are chosen into the summary. Experimental results on task 2 of DUC 2002 and task 2 of DUC 2004 demonstrate that the proposed approach outperforms existing state-of-the-art systems.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 64, "end": 69}, {"text": "based", "start": 98, "end": 103}, {"text": "approach", "start": 104, "end": 112}, {"text": "multi-document summarization", "start": 116, "end": 144}, {"text": "process", "start": 173, "end": 180}, {"text": "semantic", "start": 192, "end": 200}, {"text": "relationships", "start": 201, "end": 214}, {"text": "sentences", "start": 223, "end": 232}, {"text": "compute", "start": 243, "end": 250}, {"text": "information", "start": 251, "end": 262}, {"text": "richness", "start": 263, "end": 271}, {"text": "sentences", "start": 275, "end": 284}, {"text": "rank", "start": 296, "end": 300}, {"text": "algorithm", "start": 301, "end": 310}, {"text": "in-tra-document", "start": 329, "end": 344}, {"text": "links", "start": 345, "end": 350}, {"text": "inter-document", "start": 355, "end": 369}, {"text": "links", "start": 370, "end": 375}, {"text": "sentences", "start": 384, "end": 393}, {"text": "algorithm", "start": 404, "end": 413}, {"text": "diversity", "start": 436, "end": 445}, {"text": "sentences", "start": 457, "end": 466}, {"text": "sentences", "start": 475, "end": 484}, {"text": "information", "start": 500, "end": 511}, {"text": "richness", "start": 512, "end": 520}, {"text": "information", "start": 530, "end": 541}, {"text": "novelty", "start": 542, "end": 549}, {"text": "summary", "start": 570, "end": 577}, {"text": "Experimental", "start": 579, "end": 591}, {"text": "results", "start": 592, "end": 599}, {"text": "task", "start": 603, "end": 607}, {"text": "task", "start": 626, "end": 630}, {"text": "proposed", "start": 666, "end": 674}, {"text": "approach", "start": 675, "end": 683}, {"text": "systems", "start": 722, "end": 729}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 64, "end": 69}, "tail": {"text": "approach", "start": 104, "end": 112}}, {"head": {"text": "compute", "start": 243, "end": 250}, "tail": {"text": "algorithm", "start": 301, "end": 310}}, {"head": {"text": "algorithm", "start": 404, "end": 413}, "tail": {"text": "sentences", "start": 457, "end": 466}}, {"head": {"text": "sentences", "start": 475, "end": 484}, "tail": {"text": "summary", "start": 570, "end": 577}}]}}, "schema": []} |
| {"input": "A Hybrid Approach To Biomedical Named Entity Recognition And Semantic Role Labeling In this paper, we describe our hybrid approach to two key NLP technologies: biomedical named entity recognition (Bio-NER) and (Bio-SRL). In Bio-NER, our system successfully integrates linguistic features into the CRF framework. In addition, we employ web lexicons and template-based post-processing to further boost its performance. Through these broad linguistic features and the nature of CRF, our system outperforms state-of-the-art machine-learning-based systems, especially in the recognition of protein names (F = 78. 5%). In Bio-SRL, first, we construct a proposition bank on top of the popular biomedical GENIA treebank following the PropBank annotation scheme. We only annotate the predicate-argument structures (PAS' s) of thirty frequently used biomedical verbs (predicates) and their corresponding arguments. Second, we use our proposition bank to train a biomedical SRL system, which uses a maximum entropy (ME) machine-learning model. Thirdly, we automatically generate argument-type templates, which can be used to improve classification of biomedical argument roles. Our experimental results show that a newswire English SRL system that achieves an F-score of 86. 29% in the newswire English domain can maintain an F-score of 64. 64% when ported to the biomedical domain. By using our annotated biomedical corpus, we can increase that F-score by 22. 9%. Adding automatically generated template features further increases overall F-score by 0. 47% and adjunct (AM) F-score by 1. 57%, respectively.outputentitieskeyphrasetextpaperstartendtextapproachstartendtexttechnologiesstartendtextnamedstartendtextentitystartendtextrecognitionstartendtextsystemstartendtextlinguistic featuresstartendtextframeworkstartendtextadditionstartendtextlexiconsstartendtexttemplate-basedstartendtextpost-processingstartendtextperformancestartendtextlinguistic featuresstartendtextnaturestartendtextsystemstartendtextmachine-learning-based systemsstartendtextrecognitionstartendtextproteinstartendtextnamesstartendtextconstructstartendtextbankstartendtextschemestartendtextpredicate-argument structuresstartendtextverbsstartendtextargumentsstartendtextbankstartendtexttrainstartendtextsystemstartendtextmaximum entropystartendtextmachine-learningstartendtextmodelstartendtextgeneratestartendtextargument-typestartendtexttemplatesstartendtextimprovestartendtextclassificationstartendtextargumentstartendtextrolesstartendtextexperimentalstartendtextresultsstartendtextEnglishstartendtextsystemstartendtextEnglishstartendtextdomainstartendtextdomainstartendtextcorpusstartendtextincreasestartendtextgeneratedstartendtexttemplatestartendtextfeaturesstartendtextincreasesstartendtextadjunctstartendrelationsno_relationheadtextpaperstartendtailtextapproachstartendheadtextlinguistic featuresstartendtailtextframeworkstartendheadtextpost-processingstartendtailtextperformancestartendheadtextpredicate-argument structuresstartendtailtextverbsstartendheadtextbankstartendtailtextsystemstartendheadtexttemplatesstartendtailtextclassificationstartendschema |
| inputAutomatic Extraction Of Facts From Press Releases To Generate News Stories While complete understanding of arbitrary input text remains in the future, it is currently possible to construct natural language processing systems that provide a partial understanding of text with limited accuracy. Moreover, such systems can provide cost-effective solutions to commercially-significant business problems. This paper describes one such system: JASPER. JASPER is a fact extraction system recently developed and deployed by Carnegie Group for Reuters Ltd. JASPER uses a template-driven approach, partial understanding techniques, and heuristic procedures to extract certain key pieces of information from a limited range of text.. We believe that many significant business problems can be solved by fact extraction applications which involve locating and extracting specific, predefined types of information from a limited range of text The information extracted by such systems can be used in a variety of ways, such as filling in values in a database, generating summaries of the input text, serving as a part of the knowledge in an expert system, or feeding into another program which bases decisions on it. We expect to develop many such applications in the future using similar techniques.outputentitieskeyphrasetextunderstandingstartendtextinputstartendtexttextstartendtextconstructstartendtextnatural language processing systemsstartendtextprovidestartendtextpartialstartendtextunderstandingstartendtexttextstartendtextaccuracystartendtextsystemsstartendtextprovidestartendtextcost-effectivestartendtextsolutionsstartendtextproblemsstartendtextpaperstartendtextsystemstartendtextextraction systemstartendtextdevelopedstartendtexttemplate-drivenstartendtextapproachstartendtextpartialstartendtextunderstandingstartendtexttechniquesstartendtextproceduresstartendtextextractstartendtextinformationstartendtexttextstartendtextproblemsstartendtextsolvedstartendtextextractionstartendtextapplicationsstartendtextextractingstartendtexttypesstartendtextinformationstartendtexttextstartendtextinformationstartendtextextractedstartendtextsystemsstartendtextvarietystartendtextdatabasestartendtextgeneratingstartendtextsummariesstartendtextinputstartendtexttextstartendtextpartstartendtextknowledgestartendtextexpert systemstartendtextprogramstartendtextbasesstartendtextdecisionsstartendtextdevelopstartendtextapplicationsstartendtexttechniquesstartendrelationsno_relationheadtexttechniquesstartendtailtextextractstartendheadtextinformationstartendtailtexttextstartendheadtextinformationstartendtailtexttextstartendheadtextinformationstartendtailtextdatabasestartendheadtexttechniquesstartendtailtextapplicationsstartendschema |
| inputExploiting Sophisticated Representations For Document Retrieval \ noun group \ verb group \outputentitieskeyphrasetexttechniquesstartendtextdocument classificationstartendtextimprovementsstartendtextperformancestartendtextstandardstartendtexttermstartendtextweightingstartendtextstatisticalstartendtextassignmentstartendtextparadigmstartendtextLewisstartendtextexplanationstartendtextsolutionstartendtextdevelopedstartendtexttechniquesstartendtextappliedstartendtextmethodstartendtextcorporastartendtextstatisticalstartendtextconjunctionstartendtextstructurestartendtextmethodsstartendtextnotionsstartendtextnounstartendtextverbstartendtextrequirestartendtextcorpusstartendtextfeaturesstartendtextrepresentationstartendtextdocumentsstartendtextwordstartendtextbasedstartendtextrepresentationstartendtextunitsstartendtextpredictorsstartendtextdocumentstartendtextkeywordsstartendtextwordsstartendtextassignmentstartendtextindividualstartendtextphrasesstartendtextsumstartendtextwordstartendtextrepresentationstartendtexthigh-precisionstartendtextrulesstartendtextresultstartendtextprecisionstartendtextdevelopstartendtexttheorystartendtextrulesstartendtextresultsstartendtextstandardstartendtextstatistical modelsstartendtextrepresentationsstartendtextrule-based modelstartendtextprovidesstartendtextimprovedstartendtextperformancestartendtextrepresentationsstartendtextstatisticalstartendtextsystemsstartendtextrepresentationsstartendtextdocumentsstartendtextsupportstartendtextrepresentationsstartendtextdocument classificationstartendtextpaperstartendtextreportsstartendtextprojectstartendrelationsno_relationheadtexttechniquesstartendtailtextdocument classificationstartendheadtextrepresentationstartendtailtextdocumentsstartendheadtextrulesstartendtailtextprecisionstartendheadtextstatistical modelsstartendtailtextrule-based modelstartendheadtextrepresentationsstartendtailtextdocumentsstartendheadtextrepresentationsstartendtailtextdocument classificationstartendheadtextpaperstartendtailtextprojectstartendschema |
| inputAn Evaluation Of Lexicalization In Parsing In this paper, we evaluate a two-pass parsing strategy proposed for the so-called' lexicalized' grammar. In' lexicalized' grammars (Schabes, Abeille and Joshi, 1988), each elementary structure is systematically associated with a lexical item called A general two-pass parsing strategy for' lexicalized' grammars follows naturally. In the first stage, the parser selects a set of elementary structures associated with the lexical items in the input sentence, and in the second stage the sentence is parsed with respect to this set. We evaluate this strategy with respect to two characteristics. First, the amount of filtering on the entire grammar is evaluated: once the first pass is performed, the parser uses only a subset of the grammar. Second, we evaluate the use of non-local information: the structures selected during the first pass encode the morphological value (and therefore the position in the string) of their anchor; this enables the parser to use non-local information to guide its search. We take Lexicalized Tree Adjoining Grammars as an instance of lexicalized grammar. We illustrate the organization of the grammar. Then we show how a general Earley-type TAG parser (Schabes and Joshi, 1988) can take advantage of lexicalization. Empiricaldata show that the filtering of the grammar and the non-local information provided by the two-pass strategy improve the performance of the parser.outputentitieskeyphrasetextpaperstartendtextevaluatestartendtextparsingstartendtextstrategystartendtextproposedstartendtextstructurestartendtextlexical itemstartendtextcalledstartendtextparsingstartendtextstrategystartendtextparserstartendtextstructuresstartendtextlexical itemsstartendtextinputstartendtextsentencestartendtextsentencestartendtextrespectstartendtextevaluatestartendtextstrategystartendtextrespectstartendtextcharacteristicsstartendtextamountstartendtextevaluatedstartendtextperformedstartendtextparserstartendtextevaluatestartendtextinformationstartendtextstructuresstartendtextstringstartendtextparserstartendtextinformationstartendtextsearchstartendtextTreestartendtextinstancestartendtextorganizationstartendtextEarley-typestartendtextTAGstartendtextparserstartendtextadvantagestartendtextEmpiricaldatastartendtextinformationstartendtextprovidedstartendtextstrategystartendtextimprovestartendtextperformancestartendtextparserstartendrelationsno_relationheadtextstructuresstartendtailtextlexical itemsstartendheadtextinformationstartendtailtextparserstartendheadtextstrategystartendtailtextperformancestartendschema |
| inputContext Dependent Modeling Of Phones In Continuous Speech Using Decision Trees In a continuous speech recognition system it is important to model the context dependent variations in the pronunciations of words. In this paper we present an automatic method for modeling phonological variation using decision trees. For each phone we construct a decision tree that specifies the acoustic realization of the phone as a function of the context in which it appears. Several thousand sentences from a natural language corpus spoken by several talkers are used to construct these decision trees. Experimental results on a 5000-word vocabulary natural language speech recognition task are presented.outputentitieskeyphrasetextcontinuous speech recognition systemstartendtextmodelstartendtextcontextstartendtextvariationsstartendtextpronunciationsstartendtextwordsstartendtextpaperstartendtextautomaticstartendtextmethodstartendtextmodelingstartendtextvariationstartendtextdecision treesstartendtextphonestartendtextconstructstartendtextdecision treestartendtextrealizationstartendtextphonestartendtextfunctionstartendtextcontextstartendtextsentencesstartendtextnatural languagestartendtextcorpusstartendtextconstructstartendtextdecision treesstartendtextExperimentalstartendtextresultsstartendtextwordstartendtextvocabularystartendtextnatural languagestartendtextspeech recognitionstartendtexttaskstartendrelationsno_relationheadtextvariationsstartendtailtextpronunciationsstartendheadtextpaperstartendtailtextmethodstartendheadtextdecision treesstartendtailtextvariationstartendheadtextphonestartendtailtextdecision treestartendheadtextsentencesstartendtailtextcorpusstartendschema |
| inputExperience With A Stack Decoder-Based HMM CSR And Back-Off N-Gram Language Models Stochastic language models are more useful than non-stochastic models because they contribute more information than a simple acceptance or rejection of a word sequence. Back-ofF N-gram language models [ll] are an effective class of word based stochastic language model. The first part of this paper describes our experiences using the back-off language models in our time-synchronous decoder CSR. A bigram back-off language model was chosen for the language model to be used in the informal ATIS CSR baseline evaluation test [13, 21]. The stack decoder [2, 8, 24] is a promising control structure for a speech understanding system because it can combine constraints from both the acoustic model and a long span language model (such as a natural language processor (NLP)) into a single integrated search [17]. A copy of the Lincoln time-synchronous HMM CSR has been converted to a stack decoder controlled search with stochastic language models. The second part of this paper describes our experiences with our prototype stack decoder CSR using no grammar, the word-pair grammar, and N-gram back-off language models.outputentitieskeyphrasetextlanguage modelsstartendtextmodelsstartendtextinformationstartendtextsimplestartendtextwordstartendtextsequencestartendtextgram language modelsstartendtextclassstartendtextwordstartendtextbasedstartendtextlanguage modelstartendtextpartstartendtextpaperstartendtextexperiencesstartendtextlanguage modelsstartendtexttime-synchronousstartendtextdecoderstartendtextlanguage modelstartendtextlanguage modelstartendtextevaluationstartendtextteststartendtextdecoderstartendtextcontrolstartendtextstructurestartendtextspeech understanding systemstartendtextconstraintsstartendtextacoustic modelstartendtextspanstartendtextlanguage modelstartendtextnatural language processorstartendtextsearchstartendtexttime-synchronousstartendtextdecoderstartendtextsearchstartendtextlanguage modelsstartendtextpartstartendtextpaperstartendtextexperiencesstartendtextprototypestartendtextdecoderstartendtextword-pairstartendtextgramstartendtextlanguage modelsstartendrelationsno_relationheadtextpaperstartendtailtextexperiencesstartendheadtextlanguage modelsstartendtailtextdecoderstartendheadtextlanguage modelstartendtailtextteststartendheadtextdecoderstartendtailtextspeech understanding systemstartendheadtextpaperstartendtailtextdecoderstartendschema |
| inputHypothesizing Word Association From Untagged Text This paper reports a new method for suggesting word associations, based on a greedy algorithm that employs Chi-square statistics on joint frequencies of pairs of word groups compared against chance co-occurrence. The benefits of this new approach are: 1) we can consider even low frequency words and word pairs, and 2) word groups and word associations can be automatically generated. The method provided 87% accuracy in hypothesizing word associations for unobserved combinations of words in Japanese text.outputentitieskeyphrasetextpaperstartendtextreportsstartendtextmethodstartendtextwordstartendtextassociationsstartendtextbasedstartendtextalgorithmstartendtextstatisticsstartendtextfrequenciesstartendtextpairsstartendtextwordstartendtextco-occurrencestartendtextbenefitsstartendtextapproachstartendtextfrequencystartendtextwordsstartendtextword pairsstartendtextwordstartendtextwordstartendtextassociationsstartendtextgeneratedstartendtextmethodstartendtextprovidedstartendtextaccuracystartendtextwordstartendtextassociationsstartendtextcombinationsstartendtextwordsstartendtextJapanesestartendtexttextstartendrelationsno_relationheadtextpaperstartendtailtextmethodstartendheadtextstatisticsstartendtailtextalgorithmstartendheadtextmethodstartendtailtextaccuracystartendschema |
| inputPerceived Prosodic Boundaries And Their Phonetic Correlates This paper addresses two main questions: (a) Can listeners assign values of perceived boundary strength to the juncture between any two words? (b) If so, what is the relationship between these values and various (combinations of) suprasegmental features. Three speakers read a set of twenty utterances of varying length and complexity. A panel of nineteen listeners assigned boundary strength values to each of the 175 word boundaries in the material. Then the correlation was established between the variable strength of the perceived boundaries and three prosodie variables: melodic discontinuity, declination reset and pause. The results show that speakers may differ in their strategies of prosodie boundary marking and listeners agree in the perceptual weight they attribute to the prosodie cues.outputentitieskeyphrasetextpaperstartendtextmainstartendtextquestionsstartendtextlistenersstartendtextboundarystartendtextstrengthstartendtextwordsstartendtextrelationshipstartendtextcombinationsstartendtextfeaturesstartendtextutterancesstartendtextlengthstartendtextcomplexitystartendtextlistenersstartendtextboundarystartendtextstrengthstartendtextwordstartendtextboundariesstartendtextcorrelationstartendtextvariablestartendtextstrengthstartendtextboundariesstartendtextvariablesstartendtextresultsstartendtextstrategiesstartendtextboundarystartendtextlistenersstartendtextweightstartendtextcuesstartendrelationsno_relationheadtextboundarystartendtailtextwordsstartendheadtextboundarystartendtailtextwordstartendschema |
| inputWORDNET: A Lexical Database For English Work under this grant is intended to provide lexical resources for research on natural languages. The principal product is WordNet, a lexical database for English whose organization is inspired by current psycholinguistic theories of human lexical knowledge. Lexicalized concepts are organized by semantic relations for nouns, verbs, adjectives, and adverbs. The principal goal of the project is to upgrade WordNet and make it available to interested users. A secondary goal is to explore practical applications of WordNet; its possible use in the resolution of word senses in context (semantic disambiguation) is viewed as a necessary precursor for many other applications.outputentitieskeyphrasetextgrantstartendtextprovidestartendtextlexical resourcesstartendtextresearchstartendtextnatural languagesstartendtextproductstartendtextlexical databasestartendtextEnglishstartendtextorganizationstartendtextcurrentstartendtexttheoriesstartendtextlexical knowledgestartendtextconceptsstartendtextsemantic relationsstartendtextnounsstartendtextverbsstartendtextadjectivesstartendtextadverbsstartendtextgoalstartendtextprojectstartendtextusersstartendtextgoalstartendtextpractical applicationsstartendtextresolutionstartendtextwordstartendtextcontextstartendtextsemanticstartendtextdisambiguationstartendtextapplicationsstartendrelationsno_relationheadtextlexical resourcesstartendtailtextresearchstartendschema |
| inputIssues And Methodology For Template Design For Information Extraction The goal of Information Extraction tasks is to identify, categorize, classify, relate, and normalize specific information of interest found in free text, and to make that information available to a back-enddata base, data fusion, or other application. Adata structure referred to as a template (desiderata)outputentitieskeyphrasetextgoalstartendtextInformation ExtractionstartendtexttasksstartendtextinformationstartendtexttextstartendtextinformationstartendtextenddatastartendtextbasestartendtextdatastartendtextfusionstartendtextapplicationstartendtextAdatastartendtextstructurestartendtexttemplatestartendrelationsno_relationheadtextinformationstartendtailtexttextstartendschema |
| inputExperiments On Sentence Boundary Detection This paper explores the problem of identifying sentence boundaries in the transcriptions produced by automatic speech recognition systems. An experiment which determines the level of human performance for this task is described as well as a memory-based computational approach to the problem.outputentitieskeyphrasetextpaperstartendtextproblemstartendtextsentencestartendtextboundariesstartendtexttranscriptionsstartendtextspeech recognition systemsstartendtextexperimentstartendtextlevelstartendtextperformancestartendtexttaskstartendtextmemory-basedstartendtextcomputationalstartendtextapproachstartendtextproblemstartendrelationsno_relationheadtextboundariesstartendtailtexttranscriptionsstartendschema |
| inputMulti-Perspective Question Answering Using The OpQA Corpus We investigate techniques to support the answering of opinion-based questions. We first present the OpQA corpus of opinion questions and answers. Using the corpus, we compare and contrast the properties of fact and opinion questions and answers. Based on the disparate characteristics of opinion vs. fact answers, we argue that traditional fact-based QA approaches may have difficulty in an MPQA setting without modification. As an initial step towards the development of MPQA systems, we investigate the use of machine learning and rule-based subjectivity and opinion source filters and show that they can be used to guide MPQA systems.outputentitieskeyphrasetexttechniquesstartendtextsupportstartendtextopinion-basedstartendtextquestionsstartendtextcorpusstartendtextopinionstartendtextquestionsstartendtextcorpusstartendtextcontraststartendtextpropertiesstartendtextopinionstartendtextquestionsstartendtextBasedstartendtextcharacteristicsstartendtextopinionstartendtextapproachesstartendtextdifficultystartendtextmodificationstartendtextstepstartendtextdevelopmentstartendtextsystemsstartendtextmachine learningstartendtextrule-basedstartendtextopinionstartendtextsourcestartendtextsystemsstartendrelationsno_relationheadtextopinionstartendtailtextcorpusstartendheadtextmachine learningstartendtailtextsystemsstartendschema |
| inputOleada: User-Centered TIPSTER Technology For Language Instruction TIPSTER is an AREA sponsored program that seeks to develop methods and tools that support analysts in their efforts to filter, process, and analyze ever increasing quantities of text-based information. To this end, government sponsors, contractors, and developers are working to design an architecture specification that makes it possible for natural language processing techniques and tools, from a variety sources, to be integrated, shared, and configured by end-users. The Computing Research Laboratory (CRL) is a longtime contributor to TIPSTER. A significant portion of CRL' s research involves work on a variety of natural language processing problems, human-computer interaction, and problems associated with getting technology into the hands of end-users. CRL is using TIPSTER technology to develop OLEADA, which is an integrated set of computer tools designed to support language learners, and instructors. Further, OLEADA has been developed using a task-oriented user-centered design methodology. This paper describes the methodology used to develop OLEADA and the current system' s capabilities.outputentitieskeyphrasetextAREAstartendtextprogramstartendtextdevelopstartendtextmethodsstartendtexttoolsstartendtextsupportstartendtextanalystsstartendtexteffortsstartendtextprocessstartendtextincreasingstartendtextquantitiesstartendtexttext-basedstartendtextinformationstartendtextdevelopersstartendtextdesignstartendtextarchitecturestartendtextspecificationstartendtextnatural language processing techniquesstartendtexttoolsstartendtextvarietystartendtextsourcesstartendtextComputingstartendtextResearchstartendtextLaboratorystartendtextportionstartendtextresearchstartendtextvarietystartendtextnatural language processingstartendtextproblemsstartendtexthuman-computerstartendtextinteractionstartendtextproblemsstartendtexttechnologystartendtexthandsstartendtexttechnologystartendtextdevelopstartendtextcomputerstartendtexttoolsstartendtextdesignedstartendtextsupportstartendtextlanguagestartendtextdevelopedstartendtexttask-orientedstartendtextuser-centeredstartendtextdesignstartendtextmethodologystartendtextpaperstartendtextmethodologystartendtextdevelopstartendtextcurrentstartendtextsystemstartendtextcapabilitiesstartendrelationsno_relationheadtextpaperstartendtailtextmethodologystartendschema |
| inputThe Cornell TIPSTER Phase III Project The overall objective of the Cornell University TIPSTER Project was to improve end-user efficiency in information retrieval systems by reducing the amount of text that the user must process [1]. The project focuses on high precision IR, near-duplicate detection and context-dependent summarization. The two main foundations of the research are the latest version of the Smart system for information Retrieval and the Empire system for natural language processing. Smart is an implementation of the vector-space model of information retrieval (IR). Its earlier purpose was to provide a framework to conduct IR research but current developments will make the system easier to use by non-researcher. Empire is a research-oriented system that uses machine learning methods to quickly perform partial parsing of sentences. Cornell' s integrated approach uses both statistical and linguistic sources to first identify relationships among important terms in the query or in the text. The integrated system then uses the extracted relationships to (1) discard or reorder retrieved texts (for high-precision IR); (2) locate redundant information (for near-duplicate document detection); or (3) generate summaries. A more detailed technical description about the research can be found in the Cornell University technical paper [2].", "output": {"entities": {"keyphrase": [{"text": "objective", "start": 50, "end": 59}, {"text": "University", "start": 75, "end": 85}, {"text": "Project", "start": 94, "end": 101}, {"text": "improve", "start": 109, "end": 116}, {"text": "end-user", "start": 117, "end": 125}, {"text": "efficiency", "start": 126, "end": 136}, {"text": "information retrieval systems", "start": 140, "end": 169}, {"text": "amount", "start": 186, "end": 192}, {"text": "text", "start": 196, "end": 200}, {"text": "user", "start": 210, "end": 214}, {"text": "process", "start": 220, "end": 227}, {"text": "project", "start": 237, "end": 244}, {"text": "focuses", "start": 245, "end": 252}, {"text": "precision", "start": 261, "end": 270}, {"text": "detection", "start": 290, "end": 299}, {"text": "context-dependent", "start": 304, "end": 321}, {"text": "summarization", "start": 322, "end": 335}, {"text": "main", "start": 345, "end": 349}, {"text": "research", "start": 369, "end": 377}, {"text": "version", "start": 393, "end": 400}, {"text": "system", "start": 414, "end": 420}, {"text": "information Retrieval", "start": 425, "end": 446}, {"text": "system", "start": 462, "end": 468}, {"text": "natural language processing", "start": 473, "end": 500}, {"text": "implementation", "start": 514, "end": 528}, {"text": "vector-space", "start": 536, "end": 548}, {"text": "model", "start": 549, "end": 554}, {"text": "information retrieval", "start": 558, "end": 579}, {"text": "purpose", "start": 598, "end": 605}, {"text": "provide", "start": 613, "end": 620}, {"text": "framework", "start": 623, "end": 632}, {"text": "research", "start": 647, "end": 655}, {"text": "current", "start": 660, "end": 667}, {"text": "developments", "start": 668, "end": 680}, {"text": "system", "start": 695, "end": 701}, {"text": "non-researcher", "start": 719, "end": 733}, {"text": "research-oriented", "start": 747, "end": 764}, {"text": "system", "start": 765, "end": 771}, {"text": "machine", "start": 782, "end": 789}, {"text": "methods", "start": 799, "end": 806}, {"text": "perform", "start": 818, "end": 825}, {"text": "partial", "start": 826, "end": 833}, {"text": "parsing", "start": 834, "end": 841}, {"text": "sentences", "start": 845, "end": 854}, {"text": "approach", "start": 878, "end": 886}, {"text": "statistical", "start": 897, "end": 908}, {"text": "sources", "start": 924, "end": 931}, {"text": "relationships", "start": 950, "end": 963}, {"text": "terms", "start": 980, "end": 985}, {"text": "query", "start": 993, "end": 998}, {"text": "text", "start": 1009, "end": 1013}, {"text": "system", "start": 1030, "end": 1036}, {"text": "extracted", "start": 1051, "end": 1060}, {"text": "relationships", "start": 1061, "end": 1074}, {"text": "texts", "start": 1111, "end": 1116}, {"text": "high-precision", "start": 1122, "end": 1136}, {"text": "information", "start": 1163, "end": 1174}, {"text": "document", "start": 1195, "end": 1203}, {"text": "detection", "start": 1204, "end": 1213}, {"text": "generate", "start": 1223, "end": 1231}, {"text": "summaries", "start": 1232, "end": 1241}, {"text": "description", "start": 1269, "end": 1280}, {"text": "research", "start": 1291, "end": 1299}, {"text": "University", "start": 1328, "end": 1338}, {"text": "paper", "start": 1349, "end": 1354}]}, "relations": {"no_relation": [{"head": {"text": "Project", "start": 94, "end": 101}, "tail": {"text": "information retrieval systems", "start": 140, "end": 169}}, {"head": {"text": "parsing", "start": 834, "end": 841}, "tail": {"text": "sentences", "start": 845, "end": 854}}, {"head": {"text": "relationships", "start": 950, "end": 963}, "tail": {"text": "terms", "start": 980, "end": 985}}]}}, "schema": []} |
| {"input": "The Automatic Construction Of A Symbolic Parser Via Statistical Techniques We report on the development of a robust parsing device which aims to provide a partial explanation for child language acquisition and help in the construction of better natural language processing systems. The backbone of the new approach is the synthesis of statistical and symbolic approaches to natural language.", "output": {"entities": {"keyphrase": [{"text": "report", "start": 78, "end": 84}, {"text": "development", "start": 92, "end": 103}, {"text": "robust", "start": 109, "end": 115}, {"text": "parsing", "start": 116, "end": 123}, {"text": "device", "start": 124, "end": 130}, {"text": "provide", "start": 145, "end": 152}, {"text": "partial", "start": 155, "end": 162}, {"text": "explanation", "start": 163, "end": 174}, {"text": "language", "start": 185, "end": 193}, {"text": "acquisition", "start": 194, "end": 205}, {"text": "help", "start": 210, "end": 214}, {"text": "construction", "start": 222, "end": 234}, {"text": "natural language processing systems", "start": 245, "end": 280}, {"text": "backbone", "start": 286, "end": 294}, {"text": "approach", "start": 306, "end": 314}, {"text": "synthesis", "start": 322, "end": 331}, {"text": "statistical", "start": 335, "end": 346}, {"text": "approaches", "start": 360, "end": 370}, {"text": "natural language", "start": 374, "end": 390}]}, "relations": {"no_relation": [{"head": {"text": "device", "start": 124, "end": 130}, "tail": {"text": "natural language processing systems", "start": 245, "end": 280}}]}}, "schema": []} |
| {"input": "How Language Structures Concepts-An Outline For the past three decades, the mainstream of linguistics has focused its research agenda on the formal aspects of language, primarily syntax. By contrast, the more recent tradition of cognitive linguistics centers its research directly within the semantic stratum of language in order to observe how languages organize meaning and structure conception, and it examines the more formal stratum of language for its role in supporting these semantic functions.", "output": {"entities": {"keyphrase": [{"text": "mainstream", "start": 76, "end": 86}, {"text": "linguistics", "start": 90, "end": 101}, {"text": "focused", "start": 106, "end": 113}, {"text": "research", "start": 118, "end": 126}, {"text": "agenda", "start": 127, "end": 133}, {"text": "aspects", "start": 148, "end": 155}, {"text": "language", "start": 159, "end": 167}, {"text": "syntax", "start": 179, "end": 185}, {"text": "contrast", "start": 190, "end": 198}, {"text": "linguistics", "start": 239, "end": 250}, {"text": "centers", "start": 251, "end": 258}, {"text": "research", "start": 263, "end": 271}, {"text": "semantic", "start": 292, "end": 300}, {"text": "language", "start": 312, "end": 320}, {"text": "order", "start": 324, "end": 329}, {"text": "languages", "start": 345, "end": 354}, {"text": "structure", "start": 376, "end": 385}, {"text": "language", "start": 441, "end": 449}, {"text": "role", "start": 458, "end": 462}, {"text": "supporting", "start": 466, "end": 476}, {"text": "semantic", "start": 483, "end": 491}, {"text": "functions", "start": 492, "end": 501}]}, "relations": {}}, "schema": []} |
| {"input": "Corpus Based Statistical Generalization Tree In Rule Optimization A corpus-based statistical Generalization Tree model is described to achieve rule optimization for the information extraction task. First, the user creates specific rules for the target information from the sample articles through a training interface. Second, WordNet is applied to generalize noun entities in the specific rules. The degree of generalization is adjusted to fit the user' s needs by use of the statistical Generalization Tree model. Finally, the optimally generalized rules are applied to scan new information. The results of experiments demonstrate the applicability of our Generalization Tree method.outputentitieskeyphrasetextcorpus-basedstartendtextstatisticalstartendtextGeneralizationstartendtextTreestartendtextmodelstartendtextrulestartendtextoptimizationstartendtextextraction taskstartendtextuserstartendtextrulesstartendtexttargetstartendtextinformationstartendtextsamplestartendtexttrainingstartendtextinterfacestartendtextappliedstartendtextnounstartendtextentitiesstartendtextrulesstartendtextdegreestartendtextgeneralizationstartendtextuserstartendtextstatisticalstartendtextGeneralizationstartendtextTreestartendtextmodelstartendtextgeneralizedstartendtextrulesstartendtextappliedstartendtextinformationstartendtextresultsstartendtextexperimentsstartendtextapplicabilitystartendtextGeneralizationstartendtextTreestartendtextmethodstartendrelationsno_relationheadtextmodelstartendtailtextoptimizationstartendschema |
| inputAutomatic Lexicon Enhancement By Means Of Corpus Tagging Using specialised text corpus to automatically enhance a general lexicon is the aim of this study. Indeed, having lexicons which offer maximal cover on a specific topic is an important benefit in many applications of Automatic Speech and Natural Language Processing. The enhancement of these lexicons can be made automatic as big corpora of specialised texts are available. A syntactic tagging process, based on 3-class and 3-gram language models, allows us to automatically allocate possible syntactic categories to the Out-Of-Vocabulary (OOV) words which are found in the corpus processed. These OOV words generally occur several times in the corpus, and a number of these occurrences can be important. By taking into account all the occurrences of an OOV word in a given text as a whole, we propose here a method for automatically extracting a specialised lexicon from a text corpus which is representative of a specific topic.outputentitieskeyphrasetexttextstartendtextcorpusstartendtextlexiconstartendtextstudystartendtextlexiconsstartendtexttopicstartendtextbenefitstartendtextapplicationsstartendtextAutomaticstartendtextSpeechstartendtextNatural Language Processingstartendtextenhancementstartendtextlexiconsstartendtextautomaticstartendtextcorporastartendtexttextsstartendtextsyntacticstartendtexttaggingstartendtextprocessstartendtextbasedstartendtextclassstartendtextgram language modelsstartendtextsyntactic categoriesstartendtextVocabularystartendtextwordsstartendtextcorpusstartendtextprocessedstartendtextwordsstartendtexttimesstartendtextcorpusstartendtextnumberstartendtextoccurrencesstartendtextoccurrencesstartendtextwordstartendtexttextstartendtextproposestartendtextmethodstartendtextextractingstartendtextlexiconstartendtexttextstartendtextcorpusstartendtextrepresentativestartendtexttopicstartendrelationsno_relationheadtextcorpusstartendtailtextlexiconstartendheadtextlexiconsstartendtailtextapplicationsstartendheadtexttextsstartendtailtextcorporastartendheadtextlexiconstartendtailtexttextstartendschema |
| inputStochastic Phonological Grammars And Acceptability In foundational works of generative phonology it is claimed that subjects can reliably discriminate between possible but non-occurring words and words that could not be English. In this paper we examine the use of a probabilistic phonological parser for words to model experimentally-obtained judgements of the acceptability of a set of nonsense words. We compared various methods of scoring the goodness of the parse as a predictor of acceptability. We found that the probability of the worst part is not the best score of acceptability, indicating that classical generative phonology and Optimality Theory miss an important fact, as these approaches do not recognise a mechanism by which the frequency of well-formed parts may ameliorate the unacceptability of low-frequency parts. We argue that probabilistic generative grammars are demonstrably a more psychologically realistic model of phonological competence than standard generative phonology or Optimality Theory.outputentitieskeyphrasetextphonologystartendtextclaimedstartendtextwordsstartendtextwordsstartendtextEnglishstartendtextpaperstartendtextparserstartendtextwordsstartendtextmodelstartendtextwordsstartendtextmethodsstartendtextparsestartendtextpredictorstartendtextprobabilitystartendtextpartstartendtextphonologystartendtextTheorystartendtextapproachesstartendtextmechanismstartendtextfrequencystartendtextpartsstartendtextlow-frequencystartendtextpartsstartendtextmodelstartendtextcompetencestartendtextstandardstartendtextphonologystartendtextTheorystartendrelationsno_relationheadtextpaperstartendtailtextparserstartendheadtextfrequencystartendtailtextpartsstartendheadtextmodelstartendtailtextcompetencestartendschema |
| inputConstraints And Defaults Of Zero Pronouns In Japanese Instruction Manuals In this paper, we propose a method for anaphora resolution of zero subjects in Japanese manual sentences based on both the nature of language expressions and the ontology of ordinary instruction manuals. In instruction manuals written in Japanese, zero subjects often introduce ambiguity into sentences. In order to resolve them, we consider the property of several types of expressions including some forms of verbal phrases and some conjunctives of clauses, and so on. As the result, we have a set of constraints and defaults for zero subject resolution. We examine the precision of the constraints and defaults with real manual sentences, and we have the result that they make a good estimate with precision of over 80%.outputentitieskeyphrasetextpaperstartendtextproposestartendtextmethodstartendtextanaphora resolutionstartendtextJapanesestartendtextmanualstartendtextsentencesstartendtextbasedstartendtextnaturestartendtextlanguagestartendtextexpressionsstartendtextontologystartendtextinstructionstartendtextmanualsstartendtextinstructionstartendtextmanualsstartendtextJapanesestartendtextambiguitystartendtextsentencesstartendtextorderstartendtextpropertystartendtexttypesstartendtextexpressionsstartendtextincludingstartendtextformsstartendtextverbalstartendtextphrasesstartendtextclausesstartendtextresultstartendtextconstraintsstartendtextdefaultsstartendtextresolutionstartendtextprecisionstartendtextconstraintsstartendtextdefaultsstartendtextmanualstartendtextsentencesstartendtextresultstartendtextprecisionstartendrelationsno_relationheadtextpaperstartendtailtextmethodstartendheadtextexpressionsstartendtailtextanaphora resolutionstartendheadtextambiguitystartendtailtextsentencesstartendheadtextpropertystartendtailtextexpressionsstartendheadtextformsstartendtailtextphrasesstartendschema |
| inputAutomatic Disambiguation Of Discourse Particles In spite of their important quantitative role, discourse particles have so far been neglected in automatic speech processing for two reasons: Firstly it is not clear what they may contribute to the aims of automatic speech processing, and secondly their functions seem to vary so much that it seems difficult to identify the information relevant to such aims. The approach presented here therefore attempts to provide automatic means to distinguishing the different readings of discourse particles and to filtering out the information which can be useful for speech understanding systems, employing positional information and their role within a dialogue model of the respective domain, two types of information which are especially easy to obtain. First results indicate that discourse particles can indeed be automatically disambiguated on the basis of the model proposed.outputentitieskeyphrasetextspitestartendtextrolestartendtextdiscoursestartendtextparticlesstartendtextautomaticstartendtextspeech processingstartendtextreasonsstartendtextFirstlystartendtextautomaticstartendtextspeech processingstartendtextfunctionsstartendtextinformationstartendtextapproachstartendtextprovidestartendtextautomaticstartendtextdiscoursestartendtextparticlesstartendtextinformationstartendtextspeech understanding systemsstartendtextinformationstartendtextrolestartendtextdialoguestartendtextmodelstartendtextdomainstartendtexttypesstartendtextinformationstartendtextresultsstartendtextdiscoursestartendtextparticlesstartendtextbasisstartendtextmodelstartendtextproposedstartendrelationsno_relationheadtextspeech processingstartendtailtextparticlesstartendheadtextinformationstartendtailtextspeech understanding systemsstartendheadtextmodelstartendtailtextdomainstartendheadtextmodelstartendtailtextresultsstartendschema |
| inputDo Not Forget: Full Memory In Memory-Based Learning Of Word Pronunciation Memory-based learning, keeping full memory of learning material, appears a viable approach to learning NLP tasks, and is often superior in generalisation accuracy to eager learning approaches that abstract from learning material. Here we investigate threeoutputentitieskeyphrasetextMemory-basedstartendtextmemorystartendtextapproachstartendtextNLP tasksstartendtextaccuracystartendtextlearning approachesstartendtextabstractstartendrelationsschema |
| inputExperiments Using Stochastic Search For Text Planning Marcu has characterised an important and difficult problem in text planning: given a set of facts to convey and a set of rhetorical relations that can be used to link them together, how can one arrange this material so as to yield the best possible text? We describe experiments with a number of heuristic search methods for this task.outputentitieskeyphrasetextproblemstartendtexttextstartendtextrelationsstartendtextlinkstartendtextyieldstartendtexttextstartendtextexperimentsstartendtextnumberstartendtextsearchstartendtextmethodsstartendtexttaskstartendrelationsschema |
| inputLogical Structure And Discourse Anaphora Resolution Working within the Dynamic Quantifier Logic (DQL) framework (van den Berg 1992, 1996a, b), we claim in this paper that in every language the translation into a logical language will be such that the preference ordering of possible discourse referents for an anaphor in a sentence can be explained in terms of the scopal order of the expressions in the antecedent that introduce the discourse referents. Since the scope of terms is derived from arguments independent of any discourse theory, our account explains discourse anaphora resolution in terms of general principles of utterance semantics, from which the predictions of centering theory follow. When combined with the powerful discourse structural framework of the Linguistic Discourse Model (Polanyi (1985, 1986, 1988, 1996) Polanyi and Scha (1984), Scha and Polanyi (1988), Prfist, H., R. Scha and M. H. van den Berg, 1994; Polanyi, L. and M. H. van den Berg 1996; van den Berg, M. H. 1996b), we provide a unified account of discourse anaphora resolution.outputentitieskeyphrasetextQuantifierstartendtextLogicstartendtextframeworkstartendtextvanstartendtextdenstartendtextclaimstartendtextpaperstartendtextlanguagestartendtexttranslationstartendtextlanguagestartendtextpreferencestartendtextorderingstartendtextdiscoursestartendtextreferentsstartendtextsentencestartendtexttermsstartendtextorderstartendtextexpressionsstartendtextdiscoursestartendtextreferentsstartendtextscopestartendtexttermsstartendtextargumentsstartendtextdiscoursestartendtexttheorystartendtextdiscoursestartendtextanaphora resolutionstartendtexttermsstartendtextprinciplesstartendtextutterancestartendtextsemanticsstartendtextpredictionsstartendtextcenteringstartendtexttheorystartendtextdiscoursestartendtextstructuralstartendtextframeworkstartendtextDiscourse Modelstartendtextvanstartendtextdenstartendtextvanstartendtextdenstartendtextvanstartendtextdenstartendtextprovidestartendtextdiscoursestartendtextanaphora resolutionstartendrelationsno_relationheadtextorderstartendtailtextexpressionsstartendschema |
| inputCorpus-Based Anaphora Resolution Towards Antecedent Preference In this paper we propose a corpus-based approach to anaphora resolution combining a machine learning method and statistical information. First, a decision tree trained on an annotated corpus determines the coreference relation of a given anaphor and antecedent candidates and is utilized as a filter in order to reduce the number of potential candidates. In the second step, preference selection is achieved by taking into account the frequency information of coreferential and non-referential pairs tagged in the training corpus as well as distance features within the current discourse. Preliminary experiments concerning the resolution of Japanese pronouns in spoken-language dialogs result in a success rate of 80. 6%.outputentitieskeyphrasetextpaperstartendtextproposestartendtextcorpus-based approachstartendtextanaphora resolutionstartendtextmachinestartendtextmethodstartendtextstatisticalstartendtextinformationstartendtextdecision treestartendtexttrainedstartendtextcorpusstartendtextrelationstartendtextcandidatesstartendtextorderstartendtextnumberstartendtextcandidatesstartendtextstepstartendtextpreferencestartendtextselectionstartendtextfrequencystartendtextinformationstartendtextpairsstartendtexttaggedstartendtexttraining corpusstartendtextdistancestartendtextfeaturesstartendtextcurrentstartendtextdiscoursestartendtextexperimentsstartendtextconcerningstartendtextresolutionstartendtextJapanesestartendtextspoken-languagestartendtextdialogsstartendtextresultstartendtextsuccessstartendtextratestartendrelationsno_relationheadtextpaperstartendtailtextcorpus-based approachstartendheadtextinformationstartendtailtextpairsstartendheadtextfeaturesstartendtailtextdiscoursestartendheadtextexperimentsstartendtailtextratestartendschema |
| inputDisambiguating Toponyms In News This research is aimed at the problem of disambiguating toponyms (place names) in terms of a classification derived by merging information from two publicly available gazetteers. To establish the difficulty of the problem, we measured the degree of ambiguity, with respect to a gazetteer, for toponyms in news. We found that 67. 82% of the toponyms found in a corpus that were ambiguous in a gazetteer lacked a local discriminator in the text. Given the scarcity of human-annotated data, our method used unsuper-vised machine learning to develop disambiguation rules. Toponyms were automatically tagged with information about them found in a gazetteer. A toponym that was ambiguous in the gazetteer was automatically disambiguated based on preference heuristics. This automatically taggeddata was used to train a machine learner, which disambigu-ated toponyms in a human-annotated news corpus at 78. 5% accuracy.outputentitieskeyphrasetextresearchstartendtextproblemstartendtextdisambiguatingstartendtextplace namesstartendtexttermsstartendtextclassificationstartendtextinformationstartendtextgazetteersstartendtextdifficultystartendtextproblemstartendtextdegreestartendtextambiguitystartendtextrespectstartendtextgazetteerstartendtextnewsstartendtextcorpusstartendtextgazetteerstartendtextlackedstartendtexttextstartendtextscarcitystartendtextmethodstartendtextmachinestartendtextdevelopstartendtextdisambiguationstartendtextrulesstartendtexttaggedstartendtextinformationstartendtextgazetteerstartendtextgazetteerstartendtextbasedstartendtextpreferencestartendtexttaggeddatastartendtexttaggeddatastartendtexttrainstartendtextmachinestartendtextnewsstartendtextcorpusstartendtextaccuracystartendrelationsno_relationheadtextresearchstartendtailtextdisambiguatingstartendheadtextinformationstartendtailtextgazetteersstartendheadtextdifficultystartendtailtextproblemstartendheadtextambiguitystartendtailtextnewsstartendheadtextnewsstartendtailtextcorpusstartendschema |
| inputDisambiguation Of Morphological Structure Using A PCFG German has a productive morphology and allows the creation of complex words which are often highly ambiguous. This paper reports on the development of a head-lexicalized PCFG for the disambiguation of German morphological analyses. The grammar is trained on unla-beleddata using the Inside-Outside algorithm. The parser achieves a precision of more than 68% on difficult test data, which is 23% more than the baseline obtained by randomly choosing one of the simplest analyses. Remarkable is the fact that precision drops to 52% without lexicalization.outputentitieskeyphrasetextmorphologystartendtextcreationstartendtextcomplexstartendtextwordsstartendtextpaperstartendtextreportsstartendtextdevelopmentstartendtextdisambiguationstartendtextmorphological analysesstartendtexttrainedstartendtextbeleddatastartendtextalgorithmstartendtextparserstartendtextprecisionstartendtextteststartendtextanalysesstartendtextprecisionstartendrelationsno_relationheadtextparserstartendtailtextprecisionstartendschema |
| input\ Good Enough \ This paper proposes an end-to-end process analysis template with replicable measures to evaluate the filtering performance of a Scan-OCR-MT system. Preliminary resultsoutputentitieskeyphrasetextpaperstartendtextproposesstartendtextprocessstartendtextanalysisstartendtexttemplatestartendtextevaluatestartendtextperformancestartendtextOCR-MT systemstartendtextresultsstartendrelationsno_relationheadtextpaperstartendtailtexttemplatestartendschema |
| inputExperiments On Unsupervised Learning For Extracting Relevant Fragments From Spoken Dialog Corpus In this paper are described experiments on unsupervised learning of the domain lexicon and relevant phrase fragments from a dialog corpus. Suggested approach is based on using domain independent words for chunking and using semantical predictional power of such words for clustering and automatic extraction phrase fragments relevant to dialog topics.outputentitieskeyphrasetextpaperstartendtextexperimentsstartendtextlearningstartendtextdomainstartendtextlexiconstartendtextphrasestartendtextfragmentsstartendtextdialogstartendtextcorpusstartendtextapproachstartendtextbasedstartendtextdomainstartendtextwordsstartendtextchunkingstartendtextwordsstartendtextclusteringstartendtextautomaticstartendtextextractionstartendtextphrasestartendtextfragmentsstartendtextdialogstartendtexttopicsstartendrelationsno_relationheadtextdialogstartendtailtextcorpusstartendschema |
| inputA Paraphrase-Based Exploration Of Cohesiveness Criteria This paper proposes an empirical approach to the development of a computational model for assessing texts according to cohesiveness. We argue that the NLG technologies for the generation of structural paraphrases can be used to efficiently create what we call a cohesion-variant parallel corpus, which would serve as a good resource for empirical acquisition of cohesiveness criteria. We also present our pilot case study, in which we took a particular type of paraphrasing that separates a relative clause from a sentence. We have so far created a cohesion-variant parallel corpus containing 499 cohesive instances and 841 incohesive instances. Based on this corpus, we conducted a preliminary experiment on cohesion evaluation, obtaining encouraging results.outputentitieskeyphrasetextpaperstartendtextproposesstartendtextapproachstartendtextdevelopmentstartendtextcomputational modelstartendtexttextsstartendtexttechnologiesstartendtextgenerationstartendtextstructuralstartendtextcallstartendtextcohesion-variantstartendtextparallel corpusstartendtextresourcestartendtextacquisitionstartendtextcriteriastartendtextpilotstartendtextcase studystartendtexttypestartendtextrelativestartendtextclausestartendtextsentencestartendtextcohesion-variantstartendtextparallel corpusstartendtextinstancesstartendtextinstancesstartendtextBasedstartendtextcorpusstartendtextexperimentstartendtextcohesionstartendtextevaluationstartendtextresultsstartendrelationsno_relationheadtextpaperstartendtailtextapproachstartendheadtexttechnologiesstartendtailtextgenerationstartendheadtextinstancesstartendtailtextparallel corpusstartendheadtextexperimentstartendtailtextresultsstartendschema |
| inputTools And Resources For Tree Adjoining Grammars This paper presents a workbench for Tree Adjoining Grammars that we are currently developing. This workbench includes several tools and resources based on the markup language XML, used as a convenient language to format and exchange linguistic resources.outputentitieskeyphrasetextpaperstartendtextTreestartendtextdevelopingstartendtextincludesstartendtexttoolsstartendtextresourcesstartendtextbasedstartendtextmarkupstartendtextlanguagestartendtextlanguagestartendtextformatstartendtextlinguistic resourcesstartendrelationsschema |
| inputProcessing Comparable Corpora With Bilingual Suffix Trees We introduce Bilingual Suffix Trees (BST), adata structure that is suitable for exploiting comparable corpora. We discuss algorithms that use BSTs in order to create parallel corpora and learn translations of unseen words from comparable corpora. Starting with a small bilingual dictionary that was derived automatically from a corpus of 5. 000 parallel sentences, we have automatically extracted a corpus of 33. 926 parallel phrases of size greater than 3, and learned 9 new word translations from a comparable corpus of 1. 3M words (100. 000 sentences).outputentitieskeyphrasetextSuffixstartendtextadatastartendtextstructurestartendtextcorporastartendtextalgorithmsstartendtextorderstartendtextparallel corporastartendtexttranslationsstartendtextwordsstartendtextcorporastartendtextdictionarystartendtextcorpusstartendtextsentencesstartendtextextractedstartendtextcorpusstartendtextphrasesstartendtextsizestartendtextwordstartendtexttranslationsstartendtextcorpusstartendtextwordsstartendtextsentencesstartendrelationsno_relationheadtextsentencesstartendtailtextcorpusstartendheadtextphrasesstartendtailtextcorpusstartendheadtextwordsstartendtailtextcorpusstartendschema |
| inputBootstrapping A Multilingual Part-Of-Speech Tagger In One Person-Day This paper presents a method for bootstrapping a fine-grained, broad-coverage part-of-speech (POS) tagger in a new language using only one person-day ofdata acquisition effort. It requires only three resources, which are currently readily available in 60-100 world languages: (1) an online or hard-copy pocket-sized bilingual dictionary, (2) a basic library reference grammar, and (3) access to an existing monolingual text corpus in the language. The algorithm begins by inducing initial lexical POS distributions from English translations in a bilingual dictionary without POS tags. It handles irregular, regular and semi-regular morphology through a robust generative model using weighted Levenshtein alignments. Unsupervised induction ofgrammatical gender is performed via global modeling ofcontext-window feature agreement. Using a combination of these and other evidence sources, interactive training of context and lexical prior models are accomplished for fine-grained POS tag spaces. Experiments show high accuracy, fine-grained tag resolution with minimal new human effort.outputentitieskeyphrasetextpaperstartendtextmethodstartendtextbootstrappingstartendtextbroad-coveragestartendtextpart-of-speechstartendtextlanguagestartendtextofdatastartendtextacquisitionstartendtexteffortstartendtextrequiresstartendtextresourcesstartendtextlanguagesstartendtextdictionarystartendtextbasicstartendtextlibrarystartendtextreferencestartendtextaccessstartendtexttextstartendtextcorpusstartendtextlanguagestartendtextalgorithmstartendtextlexicalstartendtextdistributionsstartendtextEnglishstartendtexttranslationsstartendtextdictionarystartendtextPOS tagsstartendtextmorphologystartendtextrobuststartendtextgenerative modelstartendtextalignmentsstartendtextinductionstartendtextgenderstartendtextperformedstartendtextmodelingstartendtextofcontext-windowstartendtextfeaturestartendtextagreementstartendtextcombinationstartendtextevidencestartendtextsourcesstartendtexttrainingstartendtextcontextstartendtextlexicalstartendtextmodelsstartendtextPOS tagstartendtextspacesstartendtextExperimentsstartendtextaccuracystartendtexttagstartendtextresolutionstartendtexteffortstartendrelationsno_relationheadtextpaperstartendtailtextmethodstartendheadtextalignmentsstartendtailtextgenerative modelstartendheadtextExperimentsstartendtailtextaccuracystartendschema |
| inputTREQ-AL: A Word Alignment System With Limited Language Resources We provide a rather informal presentation of a prototype system for word alignment based on our previous translation equivalence approach, discuss the problems encountered in the shared-task on word-aligning of a parallel Romanian-English text, present the preliminary evaluation results and suggest further ways of improving the alignment accuracy.outputentitieskeyphrasetextprovidestartendtextpresentationstartendtextprototypestartendtextsystemstartendtextword alignmentstartendtextbasedstartendtexttranslationstartendtextequivalencestartendtextapproachstartendtextproblemsstartendtextshared-taskstartendtextword-aligningstartendtextEnglishstartendtexttextstartendtextevaluation resultsstartendtextimprovingstartendtextalignmentstartendtextaccuracystartendrelationsno_relationheadtextsystemstartendtailtextword alignmentstartendschema |
| inputWords And Pictures In The News We discuss the properties of a collection of news photos and captions, collected from the Associated Press and Reuters. Captions have a vocabulary dominated by proper names. We have implemented various text clustering algorithms to organize these items by topic, as well as an iconic matcher that identifies articles that share a picture. We have found that the special structure of captions allows us to extract some names of people actually portrayed in the image quite reliably, using a simple syntactic analysis. We have been able to build a directory of face images of individuals from this collection.outputentitieskeyphrasetextpropertiesstartendtextcollectionstartendtextnewsstartendtextvocabularystartendtextproper namesstartendtextimplementedstartendtexttextstartendtextclusteringstartendtextalgorithmsstartendtextitemsstartendtexttopicstartendtextstructurestartendtextextractstartendtextnamesstartendtextsimplestartendtextsyntactic analysisstartendtextindividualsstartendtextcollectionstartendrelationsno_relationheadtextproper namesstartendtailtextvocabularystartendheadtextextractstartendtailtextnamesstartendschema |
| inputAn Architecture For Word Learning Using Bidirectional Multimodal Structural Alignment \ The bird dove to the rock, \ to, \ to \outputentitieskeyphrasetextLearningstartendtextwordsstartendtextcontextual informationstartendtextcontextstartendtextformsstartendtextincludingstartendtextobservationsstartendtextsemanticstartendtextdomainsstartendtextcontextstartendtextwordstartendtextoutlinestartendtextarchitecturestartendtextwordstartendtextlearningstartendtextstructuralstartendtextalignmentstartendtextcontextual informationstartendtextorderstartendtextinterpretationsstartendtextunknown wordsstartendtextrelationsstartendtextsemanticstartendtextdomainstartendtextsystem-in-progressstartendtextimplementingstartendtextarchitecturestartendtextvideostartendtextsequencesstartendtextinputstartendtextexamplestartendtextsystemstartendtextvideostartendtextsequencestartendtexttreestartendtextwordsstartendtextprepositionstartendtextsystemstartendtextaspectstartendrelationsno_relationheadtextwordstartendtailtextcontextstartendheadtextinterpretationsstartendtailtextunknown wordsstartendheadtextsequencesstartendtailtextinputstartendschema |
| inputA Knowledge-Driven Approach To Text Meaning Processing \ deep \outputentitieskeyphrasetextgoalstartendtextquestionsstartendtexttextstartendtexttextstartendtexttaskstartendtextrequiresstartendtextextractingstartendtextlevelstartendtexttextstartendtextapproachstartendtextprocessingstartendtextmodelingstartendtextknowledge basestartendtextcommon-sensestartendtextexpectationsstartendtextinterpretationstartendtexttextstartendtexttextstartendtextpartsstartendtextknowledge basestartendtextpaperstartendtextinvestigationsstartendtextdevelopstartendtextapproachstartendtextmethodstartendtextprocessingstartendrelationsno_relationheadtextinterpretationstartendtailtexttextstartendheadtextpaperstartendtailtextapproachstartendheadtextmethodstartendtailtextprocessingstartendschema |
| inputAn Evolutionary Approach For Improving The Quality Of Automatic Summaries Automatic text extraction techniques have proved robust, but very often their summaries are not coherent. In this paper, we propose a new extraction method which uses local coherence as a means to improve the overall quality of automatic summaries. Two algorithms for sentence selection are proposed and evaluated on scientific documents. Evaluation showed that the method ameliorates the quality of summaries, noticeable improvements being obtained for longer summaries produced by an algorithm which selects sentences using an evolutionary algorithm.outputentitieskeyphrasetextAutomaticstartendtexttextstartendtextextractionstartendtexttechniquesstartendtextrobuststartendtextsummariesstartendtextpaperstartendtextproposestartendtextextraction methodstartendtextcoherencestartendtextimprovestartendtextqualitystartendtextautomaticstartendtextsummariesstartendtextalgorithmsstartendtextsentencestartendtextselectionstartendtextproposedstartendtextevaluatedstartendtextdocumentsstartendtextEvaluationstartendtextmethodstartendtextqualitystartendtextsummariesstartendtextimprovementsstartendtextsummariesstartendtextalgorithmstartendtextsentencesstartendtextalgorithmstartendrelationsno_relationheadtextpaperstartendtailtextextraction methodstartendheadtextalgorithmsstartendtailtextselectionstartendheadtextmethodstartendtailtextqualitystartendschema |
| inputUnsupervised Monolingual And Bilingual Word-Sense Disambiguation Of Medical Documents Using UMLS This paper describes techniques for unsupervised word sense disambiguation of English and German medical documents using UMLS. We present both monolingual techniques which rely only on the structure of UMLS, and bilingual techniques which also rely on the availability of parallel corpora. The best results are obtained using relations between terms given by UMLS, a method which achieves 74% precision, 66% coverage for English and 79% precision, 73% coverage for German on evaluation corpora and over 83% coverage over the whole corpus. The success of this technique for German shows that a lexical resource giving relations between concepts used to index an English document collection can be used for high quality disambiguation in another language.outputentitieskeyphrasetextpaperstartendtexttechniquesstartendtextword sense disambiguationstartendtextEnglishstartendtextdocumentsstartendtexttechniquesstartendtextstructurestartendtexttechniquesstartendtextavailabilitystartendtextparallel corporastartendtextresultsstartendtextrelationsstartendtexttermsstartendtextmethodstartendtextprecisionstartendtextcoveragestartendtextEnglishstartendtextprecisionstartendtextcoveragestartendtextevaluationstartendtextcorporastartendtextcoveragestartendtextcorpusstartendtextsuccessstartendtexttechniquestartendtextlexical resourcestartendtextrelationsstartendtextconceptsstartendtextindexstartendtextEnglishstartendtextdocumentstartendtextcollectionstartendtexthigh qualitystartendtextdisambiguationstartendtextlanguagestartendrelationsno_relationheadtextpaperstartendtailtexttechniquesstartendheadtextword sense disambiguationstartendtailtextdocumentsstartendheadtextstructurestartendtailtexttechniquesstartendheadtextparallel corporastartendtailtexttechniquesstartendheadtextmethodstartendtailtextprecisionstartendschema |
| inputDiscriminative Word Alignment via Alignment Matrix Modeling In this paper a new discriminative word alignment method is presented. This approach models directly the alignment matrix by a conditional random field (CRF) and so no restrictions to the alignments have to be made. Furthermore, it is easy to add features and so all available information can be used. Since the structure of the CRFs can get complex, the inference can only be done approximately and the standard algorithms had to be adapted. In addition, different methods to train the model have been developed. Using this approach the alignment quality could be improved by up to 23 percent for 3 different language pairs compared to a combination of both IBM4-alignments. Furthermore the word alignment was used to generate new phrase tables. These could improve the translation quality significantly.outputentitieskeyphrasetextpaperstartendtextword alignmentstartendtextmethodstartendtextapproachstartendtextmodelsstartendtextalignmentstartendtextmatrixstartendtextconditional random fieldstartendtextrestrictionsstartendtextalignmentsstartendtextfeaturesstartendtextinformationstartendtextstructurestartendtextcomplexstartendtextinferencestartendtextstandardstartendtextalgorithmsstartendtextadaptedstartendtextadditionstartendtextmethodsstartendtexttrainstartendtextmodelstartendtextdevelopedstartendtextapproachstartendtextalignmentstartendtextqualitystartendtextimprovedstartendtextlanguage pairsstartendtextcombinationstartendtextalignmentsstartendtextword alignmentstartendtextgeneratestartendtextphrasestartendtexttablesstartendtextimprovestartendtexttranslation qualitystartendrelationsno_relationheadtextpaperstartendtailtextmethodstartendheadtextconditional random fieldstartendtailtextapproachstartendheadtextmethodsstartendtailtexttrainstartendschema |
| inputRegularization and Search for Minimum Error Rate Training Minimum error rate training (MERT) is a widely used learning procedure for statistical machine translation models. We contrast three search strategies for MERT: Powell' s method, the variant of coordinate descent found in the Moses MERT utility, and a novel stochastic method. It is shown that the stochastic method obtains test set gains of + 0. 98 BLEU on MT03 and + 0. 61 BLEU on MT05. We also present a method for regularizing the MERT objective that achieves statistically significant gains when combined with both Powell' s method and coordinate descent.outputentitieskeyphrasetextMinimum error rate trainingstartendtextprocedurestartendtextstatistical machine translationstartendtextmodelsstartendtextcontraststartendtextsearch strategiesstartendtextmethodstartendtextvariantstartendtextutilitystartendtextmethodstartendtextmethodstartendtexttest setstartendtextgainsstartendtextmethodstartendtextobjectivestartendtextgainsstartendtextmethodstartendrelationsno_relationheadtextMinimum error rate trainingstartendtailtextstatistical machine translationstartendheadtextmethodstartendtailtextgainsstartendheadtextmethodstartendtailtextgainsstartendschema |
| inputSyntax-Driven Learning of Sub-Sentential Translation Equivalents and Translation Rules from Parsed Parallel Corpora We describe a multi-step process for automatically learning reliable sub-sentential syntactic phrases that are translation equivalents of each other and syntactic translation rules between two languages. The input to the process is a corpus of parallel sentences, word-aligned and annotated with phrase-structure parse trees. We first apply a newly developed algorithm for aligning parse-tree nodes between the two parallel trees. Next, we extract all aligned sub-sentential syntactic constituents from the parallel sentences, and create a syntax-based phrase-table. Finally, we treat the node alignments as tree decomposition points and extract from the corpus all possible synchronous parallel tree fragments. These are then converted into synchronous context-free rules. We describe the approach and analyze its application to Chinese-English paralleldata.outputentitieskeyphrasetextmulti-stepstartendtextprocessstartendtextsyntacticstartendtextphrasesstartendtexttranslationstartendtextsyntacticstartendtexttranslationstartendtextrulesstartendtextlanguagesstartendtextinputstartendtextprocessstartendtextcorpusstartendtextsentencesstartendtextword-alignedstartendtextphrase-structurestartendtextparse treesstartendtextapplystartendtextdevelopedstartendtextalgorithmstartendtextparse-treestartendtextnodesstartendtexttreesstartendtextextractstartendtextsyntacticstartendtextconstituentsstartendtextsentencesstartendtextsyntax-basedstartendtextphrase-tablestartendtextnodestartendtextalignmentsstartendtexttreestartendtextdecompositionstartendtextextractstartendtextcorpusstartendtexttreestartendtextfragmentsstartendtextcontext-freestartendtextrulesstartendtextapproachstartendtextapplicationstartendtextChinese-startendtextEnglishstartendtextparalleldatastartendrelationsno_relationheadtextsentencesstartendtailtextcorpusstartendheadtextconstituentsstartendtailtextsentencesstartendheadtextfragmentsstartendtailtextcorpusstartendschema |
| inputDetermining Verb Phrase Referents In Dialogs \ state of the world \outputentitieskeyphrasetextpaperstartendtextproblemsstartendtextinterpretationstartendtextutterancesstartendtextrelationshipstartendtextactionsstartendtextutterancestartendtexteventsstartendtextutterancesstartendtextKnowledgestartendtextlanguagestartendtextknowledgestartendtextknowledgestartendtextcurrentstartendtextsituationstartendtextproblemstartendtextactionstartendtextverbstartendtextphrasestartendtextproblemstartendtextobjectstartendtextnoun phrasestartendtextpaperstartendtextapproachstartendtextproblemstartendtextverbstartendtextphrasestartendtextreferentsstartendtextknowledgestartendtextlanguagestartendtextareastartendtextdialogstartendtextreferencesstartendtextkindsstartendtextknowledgestartendtextreferencesstartendtextactionsstartendtextalgorithmsstartendtextknowledgestartendtextdialogstartendtextutterancesstartendtexttasksstartendtextinferencesstartendtexttaskstartendtextsituationstartendtextbasedstartendtextinterpretationstartendrelationsno_relationheadtextinterpretationstartendtailtextutterancesstartendheadtextpaperstartendtailtextapproachstartendschema |
| inputAn Algorithm For Generating Quantifier Scopings The syntactic structure of a sentence often manifests quite clearly the predicate-argument structure and relations of grammatical subordination. But scope dependencies are not so transparent. As a result, many systems for representing the semantics of sentences have ignored scoping or generated scopings with mechanisms that have often been inexplicit as to the range of scopings they choose among or profligate in the scopings they allow. This paper presents, along with proofs of some of its important properties, an algorithm that generates scoped semantic forms from unscoped expressions encoding predicate-argument structure. The algorithm is not profligate as are those based on permutation of quantifiers, and it can provide a solid foundation for computational solutions where completeness is sacrificed for efficiency and heuristic efficacy.outputentitieskeyphrasetextsyntactic structurestartendtextsentencestartendtextpredicate-argument structurestartendtextrelationsstartendtextscopestartendtextdependenciesstartendtextresultstartendtextsystemsstartendtextsemanticsstartendtextsentencesstartendtextgeneratedstartendtextmechanismsstartendtextpaperstartendtextpropertiesstartendtextalgorithmstartendtextgeneratesstartendtextsemanticstartendtextformsstartendtextexpressionsstartendtextpredicate-argument structurestartendtextalgorithmstartendtextbasedstartendtextpermutationstartendtextquantifiersstartendtextprovidestartendtextcomputationalstartendtextsolutionsstartendtextcompletenessstartendtextefficiencystartendrelationsno_relationheadtextsyntactic structurestartendtailtextsentencestartendheadtextsemanticsstartendtailtextsentencesstartendheadtextpaperstartendtailtextalgorithmstartendheadtextformsstartendtailtextexpressionsstartendheadtextalgorithmstartendtailtextsolutionsstartendschema |
| inputThe Problem Of Logical-Form Equivalence \ AI-complete, \outputentitieskeyphrasetextnotestartendtextproblemstartendtextlogical-formstartendtextequivalencestartendtextproblemstartendtextnatural language generationstartendtextnotedstartendtextproblemstartendtextnatural language generation systemstartendtextproblemstartendtextgoalstartendtextgenerationstartendtextcomponentstartendtextknowledgestartendtextpaperstartendtextgenerationstartendtextcomponentstartendtextclaimedstartendtextproblemstartendtextsensestartendtextresolutionstartendtextsolutionstartendtextknowledge representationstartendtextproblemstartendtextpapersstartendtextresearchersstartendtextclaimedstartendtextsolvedstartendtextproblemstartendtextlogical-formstartendtextequivalencestartendtextpaperstartendtextreviewstartendtextproblemstartendtextaspectsstartendtextsolutionsstartendtextorderstartendtextclaimsstartendtextproblemstartendtextresolutionstartendrelationsschema |
| inputIntegrating General-Purpose And Corpus-Based Verb Classification A long-standing debate in the computational linguistic community is about the generality of lexical taxonomies. Many linguists (Nirenburg 1995; Hirst 1995) stress that taxonomies that are not language neutral, at least at the intermediate and high level, have little hope of success. On the other hand, lexicon builders who have experience of designing taxonomies for real applications claim that in sublanguages there exist very domain-dependent similarity relations. Given our experience and results, we are inclined to take the second position, but we are indeed sensitive to the theoretical motivations of the first. The problem is that the similarity relations suggested by the thematic structures of wordscompositionaloutputentitieskeyphrasetextcomputationalstartendtextcommunitystartendtextgeneralitystartendtextlexicalstartendtexttaxonomiesstartendtextlinguistsstartendtexttaxonomiesstartendtextlanguagestartendtextlevelstartendtextsuccessstartendtexthandstartendtextlexiconstartendtextexperiencestartendtextdesigningstartendtexttaxonomiesstartendtextapplicationsstartendtextclaimstartendtextdomain-dependentstartendtextsimilaritystartendtextrelationsstartendtextexperiencestartendtextresultsstartendtextmotivationsstartendtextproblemstartendtextsimilaritystartendtextrelationsstartendtextstructuresstartendrelationsno_relationheadtexttaxonomiesstartendtailtextapplicationsstartendschema |
| inputModels Of Translational Equivalence Among Words Parallel texts (bitexts) have properties that distinguish them from other kinds of paralleldata. First, most words translate to only one other word. Second, bitext correspondence is typically only partial many words in each text have no clear equivalent in the other text. This article presents methods for biasing statistical translation models to reflect these properties. Evaluation with respect to independent human judgments has confirmed that translation models biased in this fashion are significantly more accurate than a baseline knowledge-free model. This article also shows how a statistical translation model can take advantage of preexisting knowledge that might be available about particular language pairs. Even the simplest kinds of language-specific knowledge, such as the distinction between content words and function words, are shown to reliably boost translation model performance on some tasks. Statistical models that reflect knowledge about the model domain combine the best of both the rationalist and empiricist paradigms.outputentitieskeyphrasetextParallel textsstartendtextpropertiesstartendtextkindsstartendtextparalleldatastartendtextwordsstartendtexttranslatestartendtextwordstartendtextcorrespondencestartendtextpartialstartendtextwordsstartendtexttextstartendtexttextstartendtextmethodsstartendtextbiasingstartendtextstatistical translation modelsstartendtextpropertiesstartendtextEvaluationstartendtextrespectstartendtextjudgmentsstartendtexttranslation modelsstartendtextbiasedstartendtextfashionstartendtextknowledge-freestartendtextmodelstartendtextstatistical translation modelstartendtextadvantagestartendtextknowledgestartendtextlanguage pairsstartendtextkindsstartendtextlanguage-specificstartendtextknowledgestartendtextdistinctionstartendtextcontent wordsstartendtextfunction wordsstartendtexttranslation modelstartendtextperformancestartendtexttasksstartendtextStatistical modelsstartendtextknowledgestartendtextmodelstartendtextdomainstartendtextparadigmsstartendrelationsno_relationheadtextpropertiesstartendtailtextParallel textsstartendheadtextstatistical translation modelsstartendtailtextpropertiesstartendschema |
| inputHuman Variation And Lexical Choice Much natural language processing research implicitly assumes that word meanings are fixed in a language community, but in fact there is good evidence that different people probably associate slightly different meanings with words. We summarize some evidence for this claim from the literature and from an ongoing research project, and discuss its implications for natural language generation, especially for lexical choice, that is, choosing appropriate words for a generated text.outputentitieskeyphrasetextnatural language processingstartendtextresearchstartendtextword meaningsstartendtextlanguagestartendtextcommunitystartendtextevidencestartendtextwordsstartendtextevidencestartendtextclaimstartendtextliteraturestartendtextresearch projectstartendtextimplicationsstartendtextnatural language generationstartendtextlexical choicestartendtextwordsstartendtextgeneratedstartendtexttextstartendrelationsschema |
| inputStatistical Machine Translation With Scarce Resources Using Morpho-Syntactic Information In statistical machine translation, correspondences between the words in the source and the target language are learned from parallel corpora, and often little or no linguistic knowledge is used to structure the underlying models. In particular, existing statistical systems for machine translation often treat different inflected forms of the same lemma as if they were independent of one another. The bilingual trainingdata can be better exploited by explicitly taking into account the interdependencies of related inflected forms. We propose the construction of hierarchical lexicon models on the basis ofequivalence classes ofwords. In addition, we introduce sentence-level restructuring transformations which aim at the assimilation ofword order in related sentences. We have systematically investigated the amount ofbilingual trainingdata required to maintain an acceptable quality ofmachine translation. The combination ofthe suggested methods for improving translation quality in frameworks with scarce resources has been successfully tested: We were able to reduce the amount of bilingual trainingdata to less than 10% of the original corpus, while losing only1. 6% in translation quality. The improvement of the translation results is demonstrated on two German-English corpora taken from the Verbmobil task and the Nespole! task.outputentitieskeyphrasetextstatistical machine translationstartendtextcorrespondencesstartendtextwordsstartendtextsourcestartendtexttarget languagestartendtextparallel corporastartendtextlinguistic knowledgestartendtextstructurestartendtextmodelsstartendtextstatisticalstartendtextsystemsstartendtextmachine translationstartendtextformsstartendtextlemmastartendtexttrainingdatastartendtexttrainingdatastartendtextformsstartendtextproposestartendtextconstructionstartendtextlexiconstartendtextmodelsstartendtextbasisstartendtextclassesstartendtextadditionstartendtextsentence-levelstartendtexttransformationsstartendtextorderstartendtextsentencesstartendtextamountstartendtexttrainingdatastartendtexttrainingdatastartendtextrequiredstartendtextqualitystartendtexttranslationstartendtextcombinationstartendtextmethodsstartendtextimprovingstartendtexttranslation qualitystartendtextframeworksstartendtextresourcesstartendtexttestedstartendtextamountstartendtexttrainingdatastartendtexttrainingdatastartendtextcorpusstartendtexttranslation qualitystartendtextimprovementstartendtexttranslation resultsstartendtextEnglishstartendtextcorporastartendtexttaskstartendtexttaskstartendrelationsno_relationheadtextcorrespondencesstartendtailtextparallel corporastartendheadtextsystemsstartendtailtextmachine translationstartendheadtextformsstartendtailtextlemmastartendschema |
| input\ Incremental Construction And Maintenance Of Minimal Finite-State Automata, \ \ algorithm for unsorted data \ algorithm for sorted data \outputentitieskeyphrasetextalgorithmsstartendtextadditionstartendtextstringsstartendtextlanguagestartendtextremovalstartendtextstringsstartendtextalgorithmstartendtextgeneralizationstartendtextalgorithmstartendtextalgorithmsstartendtextconstructionstartendtextalgorithmstartendtextalgorithmstartendtextgeneralizedstartendtextalgorithmstartendtextalgorithmstartendtextadditionstartendrelationsschema |
| inputSemantic negotiation in dialogue: the mechanisms of alignment A key problem for models of dialogue is to explain how semantic co-ordination in dialogue is achieved and sustained. This paper presents findings from a series of Maze Task experiments which are not readily explained by the primary co-ordination mechanisms of existing models. It demonstrates that alignment in dialogue is not simply an outcome of successful interaction, but a communicative resource exploited by interlocutors in converging on a semantic model. We argue this suggests mechanisms of co-ordination in dialogue which are of relevance for a general account of how semantic co-ordination is achieved.outputentitieskeyphrasetextproblemstartendtextmodelsstartendtextdialoguestartendtextsemanticstartendtextdialoguestartendtextpaperstartendtextfindingsstartendtextseriesstartendtextTaskstartendtextexperimentsstartendtextmechanismsstartendtextmodelsstartendtextalignmentstartendtextdialoguestartendtextoutcomestartendtextinteractionstartendtextresourcestartendtextinterlocutorsstartendtextsemanticstartendtextmodelstartendtextmechanismsstartendtextdialoguestartendtextrelevancestartendtextsemanticstartendrelationsschema |
| inputA Structure-sharing Parser for Lexicalized Grammars In wide-coverage lexicalized grammars many of the elementary structures have substructures in common. This means that in conventional parsing algorithms some of the computation associated with different structures is duplicated. In this paper we describe a precompilation technique for such grammars which allows some of this computation to be shared. In our approach the elementary structures of the grammar are transformed into finite state automata which can be merged and minimised using standard algorithms, and then parsed using an automaton-based parser. We present algorithms for constructing automata from elementary structures, merging and minimising them, and string recognition and parse recovery with the resulting grammar.outputentitieskeyphrasetextwide-coveragestartendtextstructuresstartendtextsubstructuresstartendtextcommonstartendtextparsingstartendtextalgorithmsstartendtextcomputationstartendtextstructuresstartendtextpaperstartendtexttechniquestartendtextcomputationstartendtextapproachstartendtextstructuresstartendtextstandardstartendtextalgorithmsstartendtextparserstartendtextalgorithmsstartendtextconstructingstartendtextstructuresstartendtextstringstartendtextrecognitionstartendtextparsestartendtextrecoverystartendtextresultingstartendrelationsno_relationheadtextsubstructuresstartendtailtextstructuresstartendheadtextpaperstartendtailtexttechniquestartendschema |
| inputAutomatic Construction of Frame Representations for Spontaneous Speech in Unrestricted Domains This paper presents a system which automatically generates shallow semantic frame structures for conversational speech in unrestricted domains. We argue that such shallow semantic representations can indeed be generated with a minimum amount of linguistic knowledge engineering and without having to explicitly construct a semantic knowledge base. The system is designed to be robust to deal with the problems of speech dysfluencies, ungrammaticalities, and imperfect speech recognition. Initial results on speech transcripts are promising in that correct mappings could be identified in 21% of the clauses of a test set (resp. 44% of this test set where ungrammatical or verb-less clauses were removed).outputentitieskeyphrasetextpaperstartendtextsystemstartendtextgeneratesstartendtextsemanticstartendtextframestartendtextstructuresstartendtextspeechstartendtextdomainsstartendtextsemantic representationsstartendtextgeneratedstartendtextminimumstartendtextamountstartendtextlinguistic knowledgestartendtextengineeringstartendtextconstructstartendtextsemantic knowledgestartendtextbasestartendtextsystemstartendtextdesignedstartendtextrobuststartendtextdealstartendtextproblemsstartendtextspeechstartendtextspeech recognitionstartendtextresultsstartendtextspeechstartendtexttranscriptsstartendtextmappingsstartendtextclausesstartendtexttest setstartendtexttest setstartendtextverb-lessstartendtextclausesstartendrelationsno_relationheadtextpaperstartendtailtextsystemstartendschema |
| inputThe GREC Challenge 2008: Overview and Evaluation Results The grec Task at reg' 08 required participating systems to select coreference chains to the main subject of short encyclopaedic texts collected from Wikipedia. Three teams submitted a total of 6 systems, and we additionally created four baseline systems. Systems were tested automatically using a range of existing intrinsic metrics. We also evaluated systems extrinsically by applying coreference resolution tools to the outputs and measuring the success of the tools. In addition, systems were tested in a reading/comprehension experiment involving human subjects. This report describes the grec Task and the evaluation methods, gives brief descriptions of the participating systems, and presents the evaluation results.", "output": {"entities": {"keyphrase": [{"text": "Task", "start": 66, "end": 70}, {"text": "required", "start": 82, "end": 90}, {"text": "systems", "start": 105, "end": 112}, {"text": "chains", "start": 135, "end": 141}, {"text": "main", "start": 149, "end": 153}, {"text": "texts", "start": 185, "end": 190}, {"text": "systems", "start": 252, "end": 259}, {"text": "baseline systems", "start": 294, "end": 310}, {"text": "Systems", "start": 312, "end": 319}, {"text": "tested", "start": 325, "end": 331}, {"text": "metrics", "start": 382, "end": 389}, {"text": "evaluated", "start": 399, "end": 408}, {"text": "systems", "start": 409, "end": 416}, {"text": "applying", "start": 434, "end": 442}, {"text": "coreference resolution", "start": 443, "end": 465}, {"text": "tools", "start": 466, "end": 471}, {"text": "outputs", "start": 479, "end": 486}, {"text": "success", "start": 505, "end": 512}, {"text": "tools", "start": 520, "end": 525}, {"text": "addition", "start": 530, "end": 538}, {"text": "systems", "start": 540, "end": 547}, {"text": "tested", "start": 553, "end": 559}, {"text": "comprehension", "start": 573, "end": 586}, {"text": "experiment", "start": 587, "end": 597}, {"text": "report", "start": 629, "end": 635}, {"text": "Task", "start": 655, "end": 659}, {"text": "evaluation methods", "start": 668, "end": 686}, {"text": "descriptions", "start": 700, "end": 712}, {"text": "systems", "start": 734, "end": 741}, {"text": "evaluation results", "start": 760, "end": 778}]}, "relations": {"no_relation": [{"head": {"text": "report", "start": 629, "end": 635}, "tail": {"text": "evaluation methods", "start": 668, "end": 686}}]}}, "schema": []} |
| {"input": "A Hybrid Feature Set based Maximum Entropy Hindi Named Entity Recognition We describe our effort in developing a Named Entity Recognition (NER) system for Hindi using Maximum Entropy (Max-Ent) approach. We developed a NER annotated corpora for the purpose. We have tried to identify the most relevant features for Hindi NER task to enable us to develop an efficient NER from the limited corpora developed. Apart from the orthographic and collocation features, we have experimented on the efficiency of using gazetteer lists as features. We also worked on semi-automatic induction of context patterns and experimented with using these as features of the MaxEnt method. We have evaluated the performance of the system against a blind test set having 4 classes-Person, Organization, Location and Date. Our system achieved a f-value of 81. 52%.", "output": {"entities": {"keyphrase": [{"text": "effort", "start": 90, "end": 96}, {"text": "developing", "start": 100, "end": 110}, {"text": "Named", "start": 113, "end": 118}, {"text": "Entity", "start": 119, "end": 125}, {"text": "Recognition", "start": 126, "end": 137}, {"text": "system", "start": 144, "end": 150}, {"text": "Maximum Entropy", "start": 167, "end": 182}, {"text": "approach", "start": 193, "end": 201}, {"text": "developed", "start": 206, "end": 215}, {"text": "corpora", "start": 232, "end": 239}, {"text": "purpose", "start": 248, "end": 255}, {"text": "features", "start": 301, "end": 309}, {"text": "task", "start": 324, "end": 328}, {"text": "develop", "start": 345, "end": 352}, {"text": "corpora", "start": 387, "end": 394}, {"text": "developed", "start": 395, "end": 404}, {"text": "collocation", "start": 438, "end": 449}, {"text": "features", "start": 450, "end": 458}, {"text": "experimented", "start": 468, "end": 480}, {"text": "efficiency", "start": 488, "end": 498}, {"text": "gazetteer", "start": 508, "end": 517}, {"text": "lists", "start": 518, "end": 523}, {"text": "features", "start": 527, "end": 535}, {"text": "semi-automatic", "start": 555, "end": 569}, {"text": "induction", "start": 570, "end": 579}, {"text": "context", "start": 583, "end": 590}, {"text": "patterns", "start": 591, "end": 599}, {"text": "experimented", "start": 604, "end": 616}, {"text": "features", "start": 637, "end": 645}, {"text": "method", "start": 660, "end": 666}, {"text": "evaluated", "start": 676, "end": 685}, {"text": "performance", "start": 690, "end": 701}, {"text": "system", "start": 709, "end": 715}, {"text": "test set", "start": 732, "end": 740}, {"text": "classes", "start": 750, "end": 757}, {"text": "Organization", "start": 766, "end": 778}, {"text": "Location", "start": 780, "end": 788}, {"text": "Date", "start": 793, "end": 797}, {"text": "system", "start": 803, "end": 809}]}, "relations": {"no_relation": [{"head": {"text": "Maximum Entropy", "start": 167, "end": 182}, "tail": {"text": "system", "start": 144, "end": 150}}]}}, "schema": []} |
| {"input": "SVM Classification Of FrameNet Semantic Roles A Support Vector Machines (SVM) classifier of FrameNet semantic roles was implemented based on a set of new and previously used syntactic and semantic features. At Senseval 3, the system achieved a precision of 0. 807 for the restricted test and 0. 898 for the non-restricted test.", "output": {"entities": {"keyphrase": [{"text": "Support", "start": 48, "end": 55}, {"text": "Vector", "start": 56, "end": 62}, {"text": "classifier", "start": 78, "end": 88}, {"text": "semantic roles", "start": 101, "end": 115}, {"text": "implemented", "start": 120, "end": 131}, {"text": "based", "start": 132, "end": 137}, {"text": "syntactic", "start": 174, "end": 183}, {"text": "semantic features", "start": 188, "end": 205}, {"text": "system", "start": 226, "end": 232}, {"text": "precision", "start": 244, "end": 253}, {"text": "test", "start": 283, "end": 287}, {"text": "test", "start": 322, "end": 326}]}, "relations": {"no_relation": [{"head": {"text": "system", "start": 226, "end": 232}, "tail": {"text": "precision", "start": 244, "end": 253}}]}}, "schema": []} |
| {"input": "Acknowledgment Use With Synthesized And Recorded Prompts \" Acknowledgments, e. g., \" yeah \" and \" uh-huh, \" are ubiquitous in human conversation but are rarer in human-computer interaction. What interface factors might contribute to this difference? Using a simple spoken-language interface that responded to acknowledgments, we compared subjects' use of acknowledgments when the interface used recorded speech with that seen when the interface used synthesized speech. Contrary to our hypothesis, we saw a drop in the numbers of subjects using acknowledgments: subjects appeared to interpret the recorded-voice interface as signalling a more limited interface. These results were consistent for both Mexican Spanish and American English versions of the interface. \, : {: {: [{: , : 132, : 144}, {: , : 162, : 176}, {: , : 177, : 188}, {: , : 195, : 204}, {: , : 205, : 212}, {: , : 238, : 248}, {: , : 258, : 264}, {: , : 265, : 280}, {: , : 281, : 290}, {: , : 380, : 389}, {: , : 395, : 403}, {: , : 404, : 410}, {: , : 435, : 444}, {: , : 462, : 468}, {: , : 486, : 496}, {: , : 519, : 526}, {: , : 612, : 621}, {: , : 651, : 660}, {: , : 668, : 675}, {: , : 730, : 737}, {: , : 738, : 746}, {: , : 754, : 763}]}, : {: [{: {: , : 404, : 410}, : {: , : 380, : 389}}, {: {: , : 462, : 468}, : {: , : 435, : 444}}]}}, : []} |
| {: , : {: {: [{: , : 101, : 106}, {: , : 117, : 124}, {: , : 144, : 156}, {: , : 157, : 164}, {: , : 176, : 187}, {: , : 198, : 205}, {: , : 209, : 227}, {: , : 228, : 234}, {: , : 252, : 264}, {: , : 271, : 277}, {: , : 279, : 286}, {: , : 313, : 330}, {: , : 341, : 349}, {: , : 365, : 380}, {: , : 392, : 398}, {: , : 432, : 446}, {: , : 455, : 469}, {: , : 486, : 496}, {: , : 500, : 507}, {: , : 509, : 524}, {: , : 528, : 536}, {: , : 565, : 572}, {: , : 631, : 635}, {: , : 636, : 642}, {: , : 655, : 667}, {: , : 677, : 687}, {: , : 735, : 749}, {: , : 764, : 776}, {: , : 785, : 792}, {: , : 801, : 805}, {: , : 806, : 814}, {: , : 846, : 855}, {: , : 856, : 860}, {: , : 870, : 878}, {: , : 886, : 898}, {: , : 907, : 914}, {: , : 919, : 925}, {: , : 941, : 950}, {: , : 951, : 957}, {: , : 958, : 968}, {: , : 992, : 995}, {: , : 1009, : 1015}, {: , : 1016, : 1024}, {: , : 1025, : 1036}, {: , : 1044, : 1058}, {: , : 1081, : 1084}, {: , : 1098, : 1103}, {: , : 1139, : 1143}, {: , : 1178, : 1197}, {: , : 1199, : 1210}, {: , : 1214, : 1218}, {: , : 1224, : 1230}, {: , : 1242, : 1246}, {: , : 1266, : 1272}, {: , : 1314, : 1321}, {: , : 1339, : 1346}, {: , : 1395, : 1402}, {: , : 1412, : 1416}]}, : {: [{: {: , : 565, : 572}, : {: , : 636, : 642}}, {: {: , : 785, : 792}, : {: , : 806, : 814}}, {: {: , : 958, : 968}, : {: , : 1025, : 1036}}, {: {: , : 1199, : 1210}, : {: , : 1314, : 1321}}]}}, : []} |
| {: , : {: {: [{: , : 96, : 125}, {: , : 130, : 137}, {: , : 138, : 146}, {: , : 147, : 152}, {: , : 160, : 175}, {: , : 192, : 200}, {: , : 208, : 230}, {: , : 237, : 251}, {: , : 252, : 259}, {: , : 272, : 277}, {: , : 303, : 312}, {: , : 319, : 332}, {: , : 351, : 358}, {: , : 359, : 368}, {: , : 399, : 410}, {: , : 414, : 430}, {: , : 466, : 477}, {: , : 479, : 493}, {: , : 524, : 531}, {: , : 532, : 536}, {: , : 566, : 575}, {: , : 583, : 590}, {: , : 591, : 601}, {: , : 619, : 629}, {: , : 630, : 637}, {: , : 673, : 683}, {: , : 703, : 717}, {: , : 725, : 734}, {: , : 740, : 745}, {: , : 749, : 753}, {: , : 754, : 763}, {: , : 782, : 797}]}, : {: [{: {: , : 96, : 125}, : {: , : 130, : 137}}, {: {: , : 725, : 734}, : {: , : 754, : 763}}]}}, : []} |
| {: , : {: {: [{: , : 81, : 86}, {: , : 102, : 112}, {: , : 130, : 142}, {: , : 143, : 174}, {: , : 179, : 186}, {: , : 212, : 228}, {: , : 232, : 240}, {: , : 256, : 285}, {: , : 298, : 303}, {: , : 317, : 329}, {: , : 340, : 359}, {: , : 360, : 368}, {: , : 383, : 390}, {: , : 428, : 437}, {: , : 491, : 502}, {: , : 507, : 518}, {: , : 523, : 528}, {: , : 529, : 544}, {: , : 559, : 563}, {: , : 564, : 569}, {: , : 604, : 609}, {: , : 626, : 633}, {: , : 681, : 692}, {: , : 721, : 726}, {: , : 740, : 751}, {: , : 768, : 775}, {: , : 824, : 836}, {: , : 837, : 844}, {: , : 884, : 892}, {: , : 893, : 912}]}, : {: [{: {: , : 81, : 86}, : {: , : 102, : 112}}]}}, : []} |
| {: , : {: {: [{: , : 97, : 109}, {: , : 113, : 120}, {: , : 134, : 154}, {: , : 167, : 174}, {: , : 190, : 197}, {: , : 198, : 203}, {: , : 212, : 232}, {: , : 284, : 289}, {: , : 302, : 309}, {: , : 321, : 330}, {: , : 350, : 363}, {: , : 395, : 409}, {: , : 410, : 422}, {: , : 491, : 499}, {: , : 509, : 514}, {: , : 550, : 570}, {: , : 575, : 596}, {: , : 619, : 641}, {: , : 655, : 662}, {: , : 703, : 708}, {: , : 721, : 735}, {: , : 736, : 748}]}, : {: [{: {: , : 212, : 232}, : {: , : 190, : 197}}, {: {: , : 509, : 514}, : {: , : 619, : 641}}]}}, : []} |
| {: , : {: {: [{: , : 97, : 104}, {: , : 105, : 110}, {: , : 150, : 157}, {: , : 158, : 163}, {: , : 164, : 172}, {: , : 183, : 189}, {: , : 213, : 227}, {: , : 229, : 235}, {: , : 240, : 249}, {: , : 275, : 282}, {: , : 290, : 298}, {: , : 318, : 326}, {: , : 330, : 338}, {: , : 360, : 364}, {: , : 369, : 391}, {: , : 405, : 412}, {: , : 413, : 418}]}, : {: [{: {: , : 275, : 282}, : {: , : 330, : 338}}, {: {: , : 360, : 364}, : {: , : 369, : 391}}]}}, : []} |
| {: , : {: {: [{: , : 73, : 78}, {: , : 98, : 104}, {: , : 137, : 146}, {: , : 147, : 160}, {: , : 176, : 182}, {: , : 187, : 196}, {: , : 197, : 202}, {: , : 232, : 242}, {: , : 248, : 252}, {: , : 257, : 263}, {: , : 264, : 273}, {: , : 274, : 282}, {: , : 290, : 305}, {: , : 335, : 339}, {: , : 344, : 350}, {: , : 351, : 360}, {: , : 363, : 369}, {: , : 373, : 384}, {: , : 385, : 394}, {: , : 395, : 404}, {: , : 409, : 420}, {: , : 444, : 453}, {: , : 454, : 465}, {: , : 471, : 477}, {: , : 487, : 494}, {: , : 499, : 508}, {: , : 509, : 518}, {: , : 556, : 562}, {: , : 566, : 570}, {: , : 594, : 603}, {: , : 604, : 613}, {: , : 642, : 649}, {: , : 653, : 662}, {: , : 663, : 676}, {: , : 677, : 683}]}, : {: [{: {: , : 73, : 78}, : {: , : 98, : 104}}, {: {: , : 197, : 202}, : {: , : 147, : 160}}]}}, : []} |
| {: , : {: {: [{: , : 91, : 96}, {: , : 108, : 114}, {: , : 126, : 134}, {: , : 138, : 145}, {: , : 151, : 164}, {: , : 165, : 176}, {: , : 237, : 247}, {: , : 248, : 253}, {: , : 257, : 265}, {: , : 266, : 273}, {: , : 281, : 292}, {: , : 300, : 307}, {: , : 318, : 326}, {: , : 360, : 370}, {: , : 386, : 395}, {: , : 421, : 432}, {: , : 433, : 438}, {: , : 455, : 466}, {: , : 496, : 505}, {: , : 520, : 529}, {: , : 545, : 566}, {: , : 567, : 572}, {: , : 586, : 591}, {: , : 601, : 609}]}, : {: [{: {: , : 91, : 96}, : {: , : 138, : 145}}, {: {: , : 266, : 273}, : {: , : 237, : 247}}, {: {: , : 360, : 370}, : {: , : 386, : 395}}]}}, : []} |
| {: , : {: {: [{: , : 53, : 58}, {: , : 88, : 96}, {: , : 109, : 122}, {: , : 123, : 130}, {: , : 131, : 141}, {: , : 160, : 171}, {: , : 179, : 184}, {: , : 198, : 207}, {: , : 208, : 218}, {: , : 254, : 261}, {: , : 283, : 291}, {: , : 300, : 311}, {: , : 328, : 336}, {: , : 364, : 376}, {: , : 384, : 389}, {: , : 393, : 402}, {: , : 435, : 443}, {: , : 513, : 528}, {: , : 529, : 539}, {: , : 550, : 558}, {: , : 572, : 579}, {: , : 616, : 624}, {: , : 625, : 632}, {: , : 651, : 657}, {: , : 702, : 712}, {: , : 713, : 721}, {: , : 745, : 749}, {: , : 753, : 764}, {: , : 779, : 788}, {: , : 813, : 830}]}, : {: [{: {: , : 53, : 58}, : {: , : 123, : 130}}, {: {: , : 160, : 171}, : {: , : 179, : 184}}, {: {: , : 572, : 579}, : {: , : 625, : 632}}]}}, : []} |
| {: , : {: {: [{: , : 109, : 114}, {: , : 115, : 123}, {: , : 124, : 132}, {: , : 168, : 175}, {: , : 176, : 180}, {: , : 181, : 190}, {: , : 216, : 223}, {: , : 229, : 237}, {: , : 238, : 247}, {: , : 273, : 277}, {: , : 297, : 306}, {: , : 317, : 327}, {: , : 328, : 337}, {: , : 389, : 396}, {: , : 397, : 401}, {: , : 402, : 408}, {: , : 434, : 442}, {: , : 443, : 449}, {: , : 478, : 486}, {: , : 510, : 518}, {: , : 528, : 539}, {: , : 549, : 556}, {: , : 557, : 561}, {: , : 562, : 571}, {: , : 582, : 596}, {: , : 597, : 607}, {: , : 617, : 623}, {: , : 659, : 666}, {: , : 671, : 681}, {: , : 682, : 688}, {: , : 714, : 724}, {: , : 725, : 731}, {: , : 757, : 761}]}, : {: [{: {: , : 216, : 223}, : {: , : 229, : 237}}]}}, : []} |
| {: The Named Entity task consists of three subtasks (entity names, temporal expressions, number expressions). The expressions to be annotated are \ of entities (organizations, persons, locations), times (dates, times), and quantities (monetary values, percentages). For many text processing systems, such identifiers are recognized primarily using local pattern-matching techniques. The TEI (Text Encoding Initiative) Guidelines for Electronic Text Encoding and Interchange cover such identifiers (plus abbreviations) together in section 6. 4 and explain that the identifiers comprise \ The task is to identify all instances of the three types of expressions in each text in the test set and to subcategorize the expressions. The original texts contain some SGML tags already; the Named Entity task is to be performed within the text delimited by the SLUG, DATE, NWORDS, PREAMBLE, TEXT, and TRAILER tags. The system must produce a single, unambiguous output for any relevant string in the text; thus, this evaluation is not based on a view of a pipelined system architecture in which Named Entity recognition would be completely handled as a preprocess to sentence and discourse analysis. The task requires that the system recognize what a string represents, not just its superficial appearance. Sometimes, the right answer is superficially apparent, as in the case of most, if not all, NUMEX expressions, and can be obtained by local pattern-matching techniques. In other cases, the right answer is not superficially apparent, as when a single capitalized word could represent the name of a location, person, or organization, and the answer may have to be obtained using techniques that draw information from a larger context or from reference lists. The three subtasks correspond to three SGML tag elements: ENAMEX, TIMEX, and NUMEX. The subcategorization is captured by a SGML tag attribute called TYPE, which is defined to have a different set of possible values for each tag element. The markup is described in section 2, below. \, : {: {: [{: , : 56, : 61}, {: , : 62, : 68}, {: , : 69, : 73}, {: , : 102, : 108}, {: , : 109, : 114}, {: , : 125, : 136}, {: , : 138, : 144}, {: , : 145, : 156}, {: , : 163, : 174}, {: , : 204, : 215}, {: , : 221, : 229}, {: , : 231, : 244}, {: , : 255, : 264}, {: , : 267, : 272}, {: , : 274, : 279}, {: , : 281, : 286}, {: , : 293, : 303}, {: , : 322, : 333}, {: , : 345, : 360}, {: , : 361, : 368}, {: , : 375, : 386}, {: , : 424, : 440}, {: , : 441, : 451}, {: , : 462, : 466}, {: , : 476, : 486}, {: , : 488, : 498}, {: , : 514, : 518}, {: , : 555, : 566}, {: , : 573, : 586}, {: , : 600, : 607}, {: , : 634, : 645}, {: , : 665, : 673}, {: , : 741, : 745}, {: , : 747, : 752}, {: , : 754, : 759}, {: , : 764, : 771}, {: , : 803, : 813}, {: , : 840, : 844}, {: , : 848, : 854}, {: , : 861, : 869}, {: , : 891, : 896}, {: , : 918, : 922}, {: , : 1004, : 1008}, {: , : 1028, : 1037}, {: , : 1051, : 1056}, {: , : 1060, : 1071}, {: , : 1080, : 1084}, {: , : 1092, : 1100}, {: , : 1126, : 1137}, {: , : 1152, : 1157}, {: , : 1176, : 1180}, {: , : 1194, : 1199}, {: , : 1200, : 1206}, {: , : 1207, : 1211}, {: , : 1221, : 1230}, {: , : 1242, : 1246}, {: , : 1270, : 1274}, {: , : 1294, : 1298}, {: , : 1312, : 1316}, {: , : 1322, : 1328}, {: , : 1364, : 1370}, {: , : 1388, : 1394}, {: , : 1402, : 1406}, {: , : 1419, : 1429}, {: , : 1437, : 1442}, {: , : 1468, : 1487}, {: , : 1497, : 1502}, {: , : 1503, : 1509}, {: , : 1510, : 1521}, {: , : 1569, : 1577}, {: , : 1582, : 1600}, {: , : 1606, : 1610}, {: , : 1611, : 1619}, {: , : 1629, : 1635}, {: , : 1653, : 1659}, {: , : 1774, : 1778}, {: , : 1806, : 1817}, {: , : 1848, : 1864}, {: , : 1865, : 1875}, {: , : 1886, : 1891}, {: , : 1970, : 1974}, {: , : 1995, : 1999}, {: , : 2005, : 2013}, {: , : 2026, : 2038}, {: , : 2085, : 2095}, {: , : 2106, : 2117}, {: , : 2132, : 2139}, {: , : 2148, : 2157}, {: , : 2158, : 2163}, {: , : 2209, : 2212}, {: , : 2293, : 2296}, {: , : 2307, : 2313}, {: , : 2314, : 2318}, {: , : 2389, : 2392}, {: , : 2406, : 2412}, {: , : 2429, : 2436}]}, : {: [{: {: , : 441, : 451}, : {: , : 361, : 368}}, {: {: , : 665, : 673}, : {: , : 634, : 645}}, {: {: , : 1060, : 1071}, : {: , : 1080, : 1084}}, {: {: , : 1176, : 1180}, : {: , : 1152, : 1157}}, {: {: , : 1364, : 1370}, : {: , : 1388, : 1394}}, {: {: , : 1468, : 1487}, : {: , : 1419, : 1429}}, {: {: , : 1629, : 1635}, : {: , : 1606, : 1610}}, {: {: , : 2106, : 2117}, : {: , : 2158, : 2163}}]}}, : []} |
| {: , : {: {: [{: , : 66, : 71}, {: , : 86, : 94}, {: , : 109, : 118}, {: , : 119, : 129}, {: , : 137, : 151}, {: , : 152, : 164}, {: , : 177, : 188}, {: , : 195, : 207}, {: , : 215, : 224}, {: , : 226, : 230}]}, : {: [{: {: , : 86, : 94}, : {: , : 119, : 129}}]}}, : []} |
| {: , : {: {: [{: , : 73, : 79}, {: , : 121, : 128}, {: , : 150, : 154}, {: , : 165, : 178}, {: , : 179, : 187}, {: , : 243, : 255}, {: , : 261, : 268}, {: , : 280, : 285}, {: , : 308, : 314}, {: , : 352, : 356}, {: , : 379, : 388}, {: , : 419, : 424}, {: , : 430, : 436}, {: , : 444, : 451}, {: , : 452, : 464}, {: , : 468, : 476}, {: , : 511, : 520}, {: , : 533, : 541}, {: , : 542, : 552}, {: , : 569, : 578}, {: , : 626, : 638}, {: , : 648, : 653}, {: , : 671, : 685}, {: , : 708, : 714}, {: , : 722, : 729}, {: , : 742, : 756}, {: , : 770, : 777}, {: , : 803, : 808}, {: , : 823, : 829}, {: , : 833, : 842}, {: , : 845, : 851}, {: , : 852, : 872}, {: , : 877, : 881}]}, : {: [{: {: , : 280, : 285}, : {: , : 243, : 255}}, {: {: , : 379, : 388}, : {: , : 308, : 314}}, {: {: , : 569, : 578}, : {: , : 626, : 638}}, {: {: , : 648, : 653}, : {: , : 671, : 685}}]}}, : []} |
| {: , : {: {: [{: , : 55, : 61}, {: , : 72, : 80}, {: , : 89, : 96}, {: , : 118, : 127}, {: , : 175, : 183}, {: , : 184, : 190}, {: , : 210, : 220}, {: , : 225, : 233}, {: , : 245, : 252}, {: , : 269, : 276}, {: , : 300, : 309}, {: , : 336, : 353}, {: , : 408, : 419}, {: , : 453, : 459}, {: , : 463, : 474}, {: , : 507, : 512}, {: , : 558, : 568}, {: , : 601, : 608}, {: , : 624, : 630}, {: , : 656, : 666}, {: , : 685, : 692}, {: , : 693, : 700}, {: , : 701, : 712}, {: , : 731, : 739}, {: , : 753, : 763}, {: , : 778, : 789}, {: , : 801, : 808}]}, : {: [{: {: , : 72, : 80}, : {: , : 89, : 96}}, {: {: , : 245, : 252}, : {: , : 269, : 276}}, {: {: , : 731, : 739}, : {: , : 753, : 763}}, {: {: , : 778, : 789}, : {: , : 801, : 808}}]}}, : []} |
| {: , : {: {: [{: , : 85, : 90}, {: , : 120, : 126}, {: , : 142, : 162}, {: , : 182, : 194}, {: , : 212, : 218}, {: , : 219, : 224}, {: , : 230, : 236}, {: , : 240, : 247}, {: , : 255, : 260}, {: , : 261, : 267}, {: , : 282, : 289}, {: , : 299, : 306}, {: , : 307, : 316}, {: , : 322, : 334}, {: , : 335, : 340}, {: , : 345, : 353}, {: , : 365, : 375}, {: , : 376, : 386}, {: , : 406, : 416}, {: , : 417, : 423}, {: , : 440, : 447}, {: , : 468, : 481}, {: , : 504, : 519}, {: , : 529, : 541}, {: , : 529, : 541}]}, : {: [{: {: , : 85, : 90}, : {: , : 120, : 126}}, {: {: , : 182, : 194}, : {: , : 142, : 162}}, {: {: , : 230, : 236}, : {: , : 261, : 267}}, {: {: , : 335, : 340}, : {: , : 299, : 306}}, {: {: , : 417, : 423}, : {: , : 440, : 447}}, {: {: , : 504, : 519}, : {: , : 529, : 541}}]}}, : []} |
| {: , : {: {: [{: , : 59, : 66}, {: , : 77, : 83}, {: , : 125, : 129}, {: , : 157, : 164}, {: , : 172, : 185}, {: , : 190, : 201}, {: , : 216, : 225}, {: , : 231, : 239}, {: , : 240, : 254}, {: , : 265, : 269}, {: , : 298, : 304}, {: , : 309, : 317}, {: , : 354, : 368}, {: , : 393, : 399}, {: , : 440, : 446}, {: , : 463, : 473}, {: , : 474, : 480}, {: , : 485, : 499}, {: , : 500, : 509}, {: , : 538, : 548}, {: , : 556, : 563}, {: , : 564, : 574}]}, : {: [{: {: , : 77, : 83}, : {: , : 125, : 129}}]}}, : []} |
| {: , : {: {: [{: , : 68, : 75}, {: , : 92, : 100}, {: , : 130, : 139}, {: , : 166, : 171}, {: , : 177, : 184}, {: , : 207, : 214}, {: , : 259, : 266}, {: , : 267, : 274}, {: , : 275, : 285}, {: , : 290, : 297}, {: , : 329, : 334}, {: , : 338, : 343}, {: , : 351, : 361}, {: , : 376, : 383}, {: , : 398, : 403}, {: , : 404, : 411}, {: , : 437, : 444}, {: , : 454, : 459}, {: , : 475, : 486}, {: , : 490, : 494}, {: , : 500, : 505}, {: , : 527, : 533}, {: , : 553, : 559}, {: , : 573, : 589}, {: , : 590, : 600}, {: , : 620, : 630}, {: , : 645, : 651}, {: , : 666, : 671}, {: , : 672, : 681}, {: , : 694, : 699}, {: , : 704, : 710}, {: , : 711, : 717}, {: , : 727, : 734}, {: , : 745, : 757}, {: , : 758, : 772}, {: , : 773, : 783}, {: , : 789, : 796}, {: , : 804, : 815}, {: , : 865, : 873}, {: , : 909, : 914}, {: , : 920, : 927}]}, : {: [{: {: , : 454, : 459}, : {: , : 475, : 486}}]}}, : []} |
| {: , : {: {: [{: , : 93, : 98}, {: , : 102, : 117}, {: , : 118, : 123}, {: , : 154, : 161}, {: , : 182, : 187}, {: , : 199, : 209}, {: , : 241, : 253}, {: , : 241, : 253}, {: , : 271, : 279}, {: , : 311, : 316}, {: , : 339, : 346}, {: , : 372, : 381}, {: , : 382, : 391}, {: , : 473, : 483}, {: , : 487, : 494}, {: , : 518, : 528}, {: , : 536, : 543}, {: , : 547, : 557}, {: , : 558, : 565}, {: , : 566, : 578}, {: , : 600, : 613}]}, : {: [{: {: , : 93, : 98}, : {: , : 102, : 117}}, {: {: , : 241, : 253}, : {: , : 182, : 187}}, {: {: , : 518, : 528}, : {: , : 566, : 578}}]}}, : []} |
| {: , : {: {: [{: , : 62, : 71}, {: , : 72, : 84}, {: , : 103, : 107}, {: , : 133, : 143}, {: , : 144, : 150}, {: , : 158, : 177}, {: , : 183, : 188}, {: , : 234, : 243}, {: , : 244, : 254}, {: , : 296, : 304}, {: , : 314, : 334}, {: , : 335, : 345}]}, : {: [{: {: , : 72, : 84}, : {: , : 158, : 177}}, {: {: , : 335, : 345}, : {: , : 244, : 254}}]}}, : []} |
| {: , : {: {: [{: , : 95, : 100}, {: , : 112, : 118}, {: , : 141, : 150}, {: , : 154, : 158}, {: , : 159, : 167}, {: , : 168, : 181}, {: , : 204, : 209}, {: , : 217, : 223}, {: , : 236, : 242}, {: , : 248, : 256}, {: , : 257, : 263}, {: , : 279, : 288}, {: , : 301, : 306}, {: , : 319, : 325}, {: , : 326, : 332}, {: , : 357, : 364}, {: , : 378, : 386}, {: , : 414, : 419}, {: , : 436, : 445}, {: , : 449, : 456}, {: , : 504, : 523}, {: , : 528, : 544}, {: , : 534, : 544}, {: , : 565, : 574}]}, : {: [{: {: , : 95, : 100}, : {: , : 112, : 118}}, {: {: , : 534, : 544}, : {: , : 565, : 574}}]}}, : []} |
| {: , : {: {: [{: , : 101, : 109}, {: , : 110, : 121}, {: , : 142, : 148}, {: , : 157, : 179}, {: , : 180, : 184}, {: , : 201, : 221}, {: , : 223, : 246}, {: , : 256, : 264}, {: , : 265, : 273}, {: , : 302, : 319}, {: , : 325, : 333}, {: , : 348, : 356}, {: , : 422, : 427}, {: , : 431, : 439}, {: , : 456, : 460}, {: , : 471, : 478}, {: , : 488, : 493}, {: , : 497, : 504}, {: , : 508, : 516}, {: , : 559, : 567}, {: , : 572, : 594}, {: , : 641, : 649}, {: , : 650, : 658}, {: , : 683, : 691}, {: , : 706, : 714}, {: , : 715, : 726}, {: , : 727, : 738}, {: , : 744, : 754}, {: , : 758, : 765}, {: , : 758, : 765}, {: , : 785, : 792}, {: , : 793, : 798}, {: , : 799, : 819}, {: , : 835, : 857}]}, : {: [{: {: , : 265, : 273}, : {: , : 302, : 319}}, {: {: , : 488, : 493}, : {: , : 508, : 516}}, {: {: , : 559, : 567}, : {: , : 572, : 594}}, {: {: , : 799, : 819}, : {: , : 835, : 857}}]}}, : []} |
| {: , : {: {: [{: , : 110, : 119}, {: , : 120, : 132}, {: , : 133, : 139}, {: , : 149, : 154}, {: , : 176, : 182}, {: , : 199, : 205}, {: , : 218, : 223}, {: , : 234, : 243}, {: , : 244, : 256}, {: , : 260, : 267}, {: , : 281, : 286}, {: , : 314, : 318}, {: , : 331, : 341}, {: , : 351, : 357}, {: , : 358, : 363}, {: , : 367, : 383}, {: , : 402, : 410}, {: , : 415, : 423}, {: , : 415, : 423}, {: , : 429, : 440}, {: , : 457, : 467}, {: , : 468, : 476}, {: , : 484, : 490}, {: , : 494, : 500}, {: , : 524, : 532}, {: , : 534, : 537}, {: , : 566, : 574}, {: , : 575, : 585}, {: , : 589, : 599}, {: , : 589, : 599}, {: , : 609, : 616}, {: , : 619, : 625}, {: , : 626, : 640}, {: , : 641, : 647}, {: , : 681, : 689}, {: , : 699, : 709}, {: , : 723, : 736}, {: , : 744, : 750}, {: , : 786, : 793}, {: , : 794, : 804}, {: , : 805, : 815}, {: , : 827, : 834}, {: , : 838, : 845}, {: , : 875, : 882}, {: , : 890, : 896}, {: , : 897, : 903}]}, : {: [{: {: , : 281, : 286}, : {: , : 260, : 267}}, {: {: , : 367, : 383}, : {: , : 331, : 341}}, {: {: , : 484, : 490}, : {: , : 468, : 476}}]}}, : []} |
| {: , : {: {: [{: , : 41, : 54}, {: , : 58, : 68}, {: , : 83, : 109}, {: , : 114, : 119}, {: , : 127, : 138}, {: , : 170, : 178}, {: , : 186, : 193}, {: , : 231, : 237}, {: , : 243, : 247}, {: , : 275, : 296}, {: , : 297, : 305}, {: , : 310, : 317}, {: , : 326, : 338}]}, : {: [{: {: , : 83, : 109}, : {: , : 41, : 54}}]}}, : []} |
| {: , : {: {: [{: , : 81, : 85}, {: , : 102, : 113}, {: , : 140, : 147}, {: , : 174, : 180}, {: , : 201, : 209}, {: , : 261, : 273}, {: , : 298, : 308}, {: , : 312, : 319}, {: , : 338, : 349}]}, : {: [{: {: , : 298, : 308}, : {: , : 312, : 319}}]}}, : []} |
| {: , : {: {: [{: , : 71, : 77}, {: , : 78, : 84}, {: , : 94, : 101}, {: , : 133, : 142}, {: , : 195, : 208}, {: , : 209, : 218}, {: , : 252, : 261}, {: , : 262, : 268}, {: , : 273, : 281}, {: , : 316, : 322}, {: , : 377, : 382}, {: , : 390, : 394}, {: , : 410, : 416}, {: , : 421, : 429}, {: , : 435, : 441}, {: , : 449, : 457}, {: , : 458, : 465}, {: , : 525, : 533}, {: , : 534, : 545}, {: , : 567, : 576}, {: , : 592, : 602}, {: , : 626, : 637}, {: , : 638, : 649}, {: , : 666, : 681}]}, : {: [{: {: , : 262, : 268}, : {: , : 209, : 218}}, {: {: , : 638, : 649}, : {: , : 666, : 681}}]}}, : []} |
| {: For the past ten years or more, most work in the field of Natural Language Generation (NLG) has shied away from considerations regarding the processes underlying human language production. Rather, the focus has been on systems that automatically produce language usually text from non-linguistic representations, with the main objective being generation of a text that faithfully captures the meaning of those non-linguistic representations (see, e. g., Reiter and Dale' s 2000 textbook on NLG). There is, however, also a different take on NLG \" as not just competent performance by a computer but the development of a computational theory of the human capacity for language and processes that engage it \" (McDonald 1987, page 642). Guhe' s research monograph, based on his 2003 Ph. D. thesis, is firmly situated in the latter tradition. One of his main goals is to work out a computational architecture for Levelt' s (1989) psycholinguistically motivated model of language production. According to Levelt' s model, speaking involves three main activities: conceptualizing (deciding what to say), formulating (deciding how to say it), and articulating (saying it). Guhe' s book focuses on the mental activity of conceptualizing. Conceptualizing is a recalcitrant object of study, partly because of the problem of the \" initial spark \"; the decision to say something appears to be the result of volitional conscious decisions, which largely elude scientific study. Guhe avoids this problem by investigating conceptualization in settings where the main intention is already fixed: a speaker witnesses several events unfold and is instructed to describe what happens \"", "output": {"entities": {"keyphrase": [{"text": "field", "start": 134, "end": 139}, {"text": "Natural Language Generation", "start": 143, "end": 170}, {"text": "considerations", "start": 197, "end": 211}, {"text": "processes", "start": 226, "end": 235}, {"text": "human language", "start": 247, "end": 261}, {"text": "focus", "start": 286, "end": 291}, {"text": "systems", "start": 304, "end": 311}, {"text": "language", "start": 339, "end": 347}, {"text": "text", "start": 356, "end": 360}, {"text": "representations", "start": 381, "end": 396}, {"text": "main", "start": 407, "end": 411}, {"text": "objective", "start": 412, "end": 421}, {"text": "generation", "start": 428, "end": 438}, {"text": "text", "start": 444, "end": 448}, {"text": "representations", "start": 510, "end": 525}, {"text": "performance", "start": 653, "end": 664}, {"text": "computer", "start": 670, "end": 678}, {"text": "development", "start": 687, "end": 698}, {"text": "computational", "start": 704, "end": 717}, {"text": "theory", "start": 718, "end": 724}, {"text": "capacity", "start": 738, "end": 746}, {"text": "language", "start": 751, "end": 759}, {"text": "processes", "start": 764, "end": 773}, {"text": "page", "start": 807, "end": 811}, {"text": "research", "start": 826, "end": 834}, {"text": "based", "start": 846, "end": 851}, {"text": "thesis", "start": 871, "end": 877}, {"text": "main", "start": 934, "end": 938}, {"text": "goals", "start": 939, "end": 944}, {"text": "computational", "start": 962, "end": 975}, {"text": "architecture", "start": 976, "end": 988}, {"text": "model", "start": 1041, "end": 1046}, {"text": "language production", "start": 1050, "end": 1069}, {"text": "model", "start": 1094, "end": 1099}, {"text": "main", "start": 1125, "end": 1129}, {"text": "focuses", "start": 1263, "end": 1270}, {"text": "object", "start": 1348, "end": 1354}, {"text": "study", "start": 1358, "end": 1363}, {"text": "problem", "start": 1387, "end": 1394}, {"text": "decision", "start": 1425, "end": 1433}, {"text": "result", "start": 1469, "end": 1475}, {"text": "decisions", "start": 1500, "end": 1509}, {"text": "study", "start": 1542, "end": 1547}, {"text": "problem", "start": 1566, "end": 1573}, {"text": "main", "start": 1631, "end": 1635}, {"text": "events", "start": 1692, "end": 1698}]}, "relations": {"no_relation": [{"head": {"text": "model", "start": 1041, "end": 1046}, "tail": {"text": "language production", "start": 1050, "end": 1069}}]}}, "schema": []} |
| {"input": "First ideas of user-adapted views of lexicographicdata exemplified on OWID and elexiko This paper is a project report of the lexicographic Internet portal OWID, an Online Vocabulary Information System of German which is being built at the Institute of German Language in Mannheim (IDS). Overall, the contents of the portal and its technical approaches will be presented. The lexical database is structured in a granular way which allows to extend possible search options for lexicographers. Against the background of current research on using electronic dictionaries, the project OWID is also working on first ideas of user-adapted access and user-adapted views of the lexicographicdata. Due to the fact that the portal OWID comprises dictionaries which are available online it is possible to change the design and functions of the website easily (in comparison to printed dictionaries). Ideas of implementing user-adapted views of the lexicographicdata will be demonstrated by using an example taken from one of the dictionaries of the portal, namely elexiko.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 92, "end": 97}, {"text": "project", "start": 103, "end": 110}, {"text": "report", "start": 111, "end": 117}, {"text": "Vocabulary", "start": 171, "end": 181}, {"text": "Information System", "start": 182, "end": 200}, {"text": "Language", "start": 259, "end": 267}, {"text": "approaches", "start": 341, "end": 351}, {"text": "lexical database", "start": 375, "end": 391}, {"text": "structured", "start": 395, "end": 405}, {"text": "search", "start": 456, "end": 462}, {"text": "options", "start": 463, "end": 470}, {"text": "background", "start": 503, "end": 513}, {"text": "current", "start": 517, "end": 524}, {"text": "research", "start": 525, "end": 533}, {"text": "electronic dictionaries", "start": 543, "end": 566}, {"text": "project", "start": 572, "end": 579}, {"text": "user-adapted", "start": 619, "end": 631}, {"text": "access", "start": 632, "end": 638}, {"text": "user-adapted", "start": 643, "end": 655}, {"text": "lexicographicdata", "start": 669, "end": 686}, {"text": "Due", "start": 688, "end": 691}, {"text": "dictionaries", "start": 735, "end": 747}, {"text": "design", "start": 804, "end": 810}, {"text": "functions", "start": 815, "end": 824}, {"text": "comparison", "start": 851, "end": 861}, {"text": "dictionaries", "start": 873, "end": 885}, {"text": "implementing", "start": 897, "end": 909}, {"text": "user-adapted", "start": 910, "end": 922}, {"text": "lexicographicdata", "start": 936, "end": 953}, {"text": "example", "start": 987, "end": 994}, {"text": "dictionaries", "start": 1017, "end": 1029}]}, "relations": {"no_relation": [{"head": {"text": "example", "start": 987, "end": 994}, "tail": {"text": "dictionaries", "start": 1017, "end": 1029}}]}}, "schema": []} |
| {"input": "Probabilistic Model for Syntactic and Semantic Dependency Parsing This paper proposes a novel method to analyze syntactic dependencies and label semantic dependencies around both the verbal predicates and the nouns. In this method, a probabilistic model is designed to obtain a global optimal result. Moreover, a predicate identification model and a disambiguation model are proposed to label predicates and their senses. The experimental results obtained on the wsj and brown test sets show that our system obtains 77% of labeled macro F1 score for the whole task, 84. 47% of labeled attachment score for syntactic dependency task, and 69. 45% of labeled F1 score for semantic dependency task.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 71, "end": 76}, {"text": "proposes", "start": 77, "end": 85}, {"text": "method", "start": 94, "end": 100}, {"text": "syntactic", "start": 112, "end": 121}, {"text": "dependencies", "start": 122, "end": 134}, {"text": "semantic", "start": 145, "end": 153}, {"text": "dependencies", "start": 154, "end": 166}, {"text": "verbal", "start": 183, "end": 189}, {"text": "nouns", "start": 209, "end": 214}, {"text": "method", "start": 224, "end": 230}, {"text": "probabilistic model", "start": 234, "end": 253}, {"text": "designed", "start": 257, "end": 265}, {"text": "optimal", "start": 285, "end": 292}, {"text": "result", "start": 293, "end": 299}, {"text": "identification", "start": 323, "end": 337}, {"text": "model", "start": 338, "end": 343}, {"text": "disambiguation", "start": 350, "end": 364}, {"text": "model", "start": 365, "end": 370}, {"text": "proposed", "start": 375, "end": 383}, {"text": "experimental", "start": 426, "end": 438}, {"text": "results", "start": 439, "end": 446}, {"text": "test sets", "start": 477, "end": 486}, {"text": "system", "start": 501, "end": 507}, {"text": "task", "start": 560, "end": 564}, {"text": "attachment", "start": 585, "end": 595}, {"text": "syntactic", "start": 606, "end": 615}, {"text": "dependency", "start": 616, "end": 626}, {"text": "task", "start": 627, "end": 631}, {"text": "semantic", "start": 669, "end": 677}, {"text": "dependency", "start": 678, "end": 688}, {"text": "task", "start": 689, "end": 693}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 71, "end": 76}, "tail": {"text": "method", "start": 94, "end": 100}}, {"head": {"text": "probabilistic model", "start": 234, "end": 253}, "tail": {"text": "result", "start": 293, "end": 299}}]}}, "schema": []} |
| {"input": "Are Very Large N-Best Lists Useful for SMT? This paper describes an efficient method to extract large n-best lists from a word graph produced by a statistical machine translation system. The extraction is based on the k shortest paths algorithm which is efficient even for very large k. We show that, although we can generate large amounts of distinct translation hypotheses, these numerous candidates are not able to significantly improve overall system performance. We conclude that large n-best lists would benefit from better discriminating models.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 49, "end": 54}, {"text": "method", "start": 78, "end": 84}, {"text": "extract", "start": 88, "end": 95}, {"text": "lists", "start": 109, "end": 114}, {"text": "word", "start": 122, "end": 126}, {"text": "statistical machine translation system", "start": 147, "end": 185}, {"text": "extraction", "start": 191, "end": 201}, {"text": "based", "start": 205, "end": 210}, {"text": "paths", "start": 229, "end": 234}, {"text": "algorithm", "start": 235, "end": 244}, {"text": "generate", "start": 317, "end": 325}, {"text": "amounts", "start": 332, "end": 339}, {"text": "translation", "start": 352, "end": 363}, {"text": "hypotheses", "start": 364, "end": 374}, {"text": "candidates", "start": 391, "end": 401}, {"text": "improve", "start": 432, "end": 439}, {"text": "system", "start": 448, "end": 454}, {"text": "performance", "start": 455, "end": 466}, {"text": "lists", "start": 498, "end": 503}, {"text": "benefit", "start": 510, "end": 517}, {"text": "models", "start": 545, "end": 551}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 49, "end": 54}, "tail": {"text": "method", "start": 78, "end": 84}}, {"head": {"text": "algorithm", "start": 235, "end": 244}, "tail": {"text": "extraction", "start": 191, "end": 201}}]}}, "schema": []} |
| {"input": "A Word-Order Database For Testing Computational Models Of Language Acquisition An investment of effort over the last two years has begun to produce a wealth ofdata concerning computational psycholin-guistic models of syntax acquisition. Thedata is generated by running simulations on a recently completed database of word order patterns from over 3, 000 abstract languages. This article presents the design of the database which contains sentence patterns, grammars and derivations that can be used to test acquisition models from widely divergent paradigms. The domain is generated from grammars that are linguistically motivated by current syntactic theory and the sentence patterns have been validated as psychologically/developmen-tally plausible by checking their frequency of occurrence in corpora of child-directed speech. A small case-study simulation is also presented.", "output": {"entities": {"keyphrase": [{"text": "effort", "start": 96, "end": 102}, {"text": "ofdata", "start": 157, "end": 163}, {"text": "concerning", "start": 164, "end": 174}, {"text": "computational", "start": 175, "end": 188}, {"text": "models", "start": 207, "end": 213}, {"text": "syntax", "start": 217, "end": 223}, {"text": "acquisition", "start": 224, "end": 235}, {"text": "Thedata", "start": 237, "end": 244}, {"text": "generated", "start": 248, "end": 257}, {"text": "simulations", "start": 269, "end": 280}, {"text": "database", "start": 305, "end": 313}, {"text": "word", "start": 317, "end": 321}, {"text": "order", "start": 322, "end": 327}, {"text": "patterns", "start": 328, "end": 336}, {"text": "abstract", "start": 354, "end": 362}, {"text": "languages", "start": 363, "end": 372}, {"text": "design", "start": 400, "end": 406}, {"text": "database", "start": 414, "end": 422}, {"text": "sentence", "start": 438, "end": 446}, {"text": "patterns", "start": 447, "end": 455}, {"text": "derivations", "start": 470, "end": 481}, {"text": "test", "start": 502, "end": 506}, {"text": "acquisition", "start": 507, "end": 518}, {"text": "models", "start": 519, "end": 525}, {"text": "paradigms", "start": 548, "end": 557}, {"text": "domain", "start": 563, "end": 569}, {"text": "generated", "start": 573, "end": 582}, {"text": "current", "start": 634, "end": 641}, {"text": "syntactic", "start": 642, "end": 651}, {"text": "theory", "start": 652, "end": 658}, {"text": "sentence", "start": 667, "end": 675}, {"text": "patterns", "start": 676, "end": 684}, {"text": "checking", "start": 754, "end": 762}, {"text": "frequency", "start": 769, "end": 778}, {"text": "occurrence", "start": 782, "end": 792}, {"text": "corpora", "start": 796, "end": 803}, {"text": "speech", "start": 822, "end": 828}, {"text": "case-study", "start": 838, "end": 848}, {"text": "simulation", "start": 849, "end": 859}]}, "relations": {"no_relation": [{"head": {"text": "patterns", "start": 328, "end": 336}, "tail": {"text": "database", "start": 305, "end": 313}}, {"head": {"text": "patterns", "start": 447, "end": 455}, "tail": {"text": "database", "start": 414, "end": 422}}, {"head": {"text": "speech", "start": 822, "end": 828}, "tail": {"text": "corpora", "start": 796, "end": 803}}]}}, "schema": []} |
| {"input": "Deverbal Compound Noun Analysis Based On Lexical Conceptual Structure This paper proposes a principled approach for analysis of semantic relations between constituents in compound nouns based on lexical semantic structure. One of the difficulties of compound noun analysis is that the mechanisms governing the decision system of semantic relations and the representation method of semantic relations associated with lexical and contextual meaning are not obvious. The aim of our research is to clarify how lexical semantics contribute to the relations in compound nouns since such nouns are very productive and are supposed to be governed by systematic mechanisms. The results of applying our approach to the analysis of noun-deverbal compounds in Japanese and English show that lexical conceptual structure contributes to the re-strictional rules in compounds.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 75, "end": 80}, {"text": "proposes", "start": 81, "end": 89}, {"text": "approach", "start": 103, "end": 111}, {"text": "analysis", "start": 116, "end": 124}, {"text": "semantic relations", "start": 128, "end": 146}, {"text": "constituents", "start": 155, "end": 167}, {"text": "nouns", "start": 180, "end": 185}, {"text": "based", "start": 186, "end": 191}, {"text": "lexical", "start": 195, "end": 202}, {"text": "semantic structure", "start": 203, "end": 221}, {"text": "difficulties", "start": 234, "end": 246}, {"text": "noun", "start": 259, "end": 263}, {"text": "analysis", "start": 264, "end": 272}, {"text": "mechanisms", "start": 285, "end": 295}, {"text": "decision", "start": 310, "end": 318}, {"text": "system", "start": 319, "end": 325}, {"text": "semantic relations", "start": 329, "end": 347}, {"text": "representation", "start": 356, "end": 370}, {"text": "method", "start": 371, "end": 377}, {"text": "semantic relations", "start": 381, "end": 399}, {"text": "lexical", "start": 416, "end": 423}, {"text": "research", "start": 479, "end": 487}, {"text": "lexical semantics", "start": 506, "end": 523}, {"text": "relations", "start": 542, "end": 551}, {"text": "nouns", "start": 564, "end": 569}, {"text": "nouns", "start": 581, "end": 586}, {"text": "mechanisms", "start": 653, "end": 663}, {"text": "results", "start": 669, "end": 676}, {"text": "applying", "start": 680, "end": 688}, {"text": "approach", "start": 693, "end": 701}, {"text": "analysis", "start": 709, "end": 717}, {"text": "noun-deverbal", "start": 721, "end": 734}, {"text": "Japanese", "start": 748, "end": 756}, {"text": "English", "start": 761, "end": 768}, {"text": "lexical", "start": 779, "end": 786}, {"text": "conceptual structure", "start": 787, "end": 807}, {"text": "rules", "start": 842, "end": 847}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 75, "end": 80}, "tail": {"text": "approach", "start": 103, "end": 111}}, {"head": {"text": "analysis", "start": 116, "end": 124}, "tail": {"text": "semantic relations", "start": 128, "end": 146}}]}}, "schema": []} |
| {"input": "Computing Locally Coherent Discourses We present the first algorithm that computes optimal orderings of sentences into a locally coherent discourse. The algorithm runs very efficiently on a variety of coherence measures from the literature. We also show that the discourse ordering problem is NP-complete and cannot be approximated.", "output": {"entities": {"keyphrase": [{"text": "algorithm", "start": 59, "end": 68}, {"text": "computes", "start": 74, "end": 82}, {"text": "optimal", "start": 83, "end": 90}, {"text": "sentences", "start": 104, "end": 113}, {"text": "discourse", "start": 138, "end": 147}, {"text": "algorithm", "start": 153, "end": 162}, {"text": "variety", "start": 190, "end": 197}, {"text": "coherence", "start": 201, "end": 210}, {"text": "literature", "start": 229, "end": 239}, {"text": "discourse", "start": 263, "end": 272}, {"text": "ordering", "start": 273, "end": 281}, {"text": "problem", "start": 282, "end": 289}]}, "relations": {"no_relation": [{"head": {"text": "algorithm", "start": 59, "end": 68}, "tail": {"text": "sentences", "start": 104, "end": 113}}]}}, "schema": []} |
| {"input": "Multi-Engine Machine Translation With Voted Language Model The paper describes a particular approach to multi-engine machine translation (MEMT), where we make use of voted language models to selectively combine translation outputs from multiple off-the-shelf MT systems. Experiments are done using large corpora from three distinct domains. The study found that the use of voted language models leads to an improved performance of MEMT systems.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 63, "end": 68}, {"text": "approach", "start": 92, "end": 100}, {"text": "multi-engine", "start": 104, "end": 116}, {"text": "machine translation", "start": 117, "end": 136}, {"text": "language models", "start": 172, "end": 187}, {"text": "translation", "start": 211, "end": 222}, {"text": "outputs", "start": 223, "end": 230}, {"text": "MT systems", "start": 259, "end": 269}, {"text": "Experiments", "start": 271, "end": 282}, {"text": "corpora", "start": 304, "end": 311}, {"text": "domains", "start": 332, "end": 339}, {"text": "study", "start": 345, "end": 350}, {"text": "language models", "start": 379, "end": 394}, {"text": "improved", "start": 407, "end": 415}, {"text": "performance", "start": 416, "end": 427}, {"text": "systems", "start": 436, "end": 443}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 63, "end": 68}, "tail": {"text": "approach", "start": 92, "end": 100}}, {"head": {"text": "language models", "start": 172, "end": 187}, "tail": {"text": "machine translation", "start": 117, "end": 136}}, {"head": {"text": "corpora", "start": 304, "end": 311}, "tail": {"text": "Experiments", "start": 271, "end": 282}}, {"head": {"text": "language models", "start": 379, "end": 394}, "tail": {"text": "performance", "start": 416, "end": 427}}]}}, "schema": []} |
| {"input": "A Unified Framework For Automatic Evaluation Using N-Gram Co-Occurrence Statistics In this paper we propose a unified framework for automatic evaluation of NLP applications using N-gram co-occurrence statistics. The automatic evaluation metrics proposed to date for Machine Translation and Automatic Summarization are particular instances from the family of metrics we propose. We show that different members of the same family of metrics explain best the variations obtained with human evaluations, according to the application being evaluated (Machine Translation, Automatic Summarization, and Automatic Question Answering) and the evaluation guidelines used by humans for evaluating such applications.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 91, "end": 96}, {"text": "propose", "start": 100, "end": 107}, {"text": "framework", "start": 118, "end": 127}, {"text": "automatic evaluation", "start": 132, "end": 152}, {"text": "NLP applications", "start": 156, "end": 172}, {"text": "gram", "start": 181, "end": 185}, {"text": "co-occurrence", "start": 186, "end": 199}, {"text": "statistics", "start": 200, "end": 210}, {"text": "evaluation metrics", "start": 226, "end": 244}, {"text": "proposed", "start": 245, "end": 253}, {"text": "date", "start": 257, "end": 261}, {"text": "Machine Translation", "start": 266, "end": 285}, {"text": "Automatic Summarization", "start": 290, "end": 313}, {"text": "instances", "start": 329, "end": 338}, {"text": "metrics", "start": 358, "end": 365}, {"text": "propose", "start": 369, "end": 376}, {"text": "metrics", "start": 431, "end": 438}, {"text": "variations", "start": 456, "end": 466}, {"text": "evaluations", "start": 487, "end": 498}, {"text": "application", "start": 517, "end": 528}, {"text": "evaluated", "start": 535, "end": 544}, {"text": "Machine Translation", "start": 546, "end": 565}, {"text": "Automatic Summarization", "start": 567, "end": 590}, {"text": "Automatic", "start": 596, "end": 605}, {"text": "Question Answering", "start": 606, "end": 624}, {"text": "evaluation", "start": 634, "end": 644}, {"text": "guidelines", "start": 645, "end": 655}, {"text": "evaluating", "start": 675, "end": 685}, {"text": "applications", "start": 691, "end": 703}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 91, "end": 96}, "tail": {"text": "framework", "start": 118, "end": 127}}, {"head": {"text": "statistics", "start": 200, "end": 210}, "tail": {"text": "NLP applications", "start": 156, "end": 172}}, {"head": {"text": "evaluation metrics", "start": 226, "end": 244}, "tail": {"text": "Machine Translation", "start": 266, "end": 285}}, {"head": {"text": "variations", "start": 456, "end": 466}, "tail": {"text": "evaluations", "start": 487, "end": 498}}]}}, "schema": []} |
| {"input": "MT Evaluation: Human-Like Vs. Human Acceptable We present a comparative study on Machine Translation Evaluation according to two different criteria: Human Likeness and Human Acceptability. We provide empirical evidence that there is a relationship between these two kinds of evaluation: Human Likeness implies Human Acceptability but the reverse is not true. From the point of view of automatic evaluation this implies that metrics based on Human Likeness are more reliable for system tuning. Our results also show that current evaluation metrics are not always able to distinguish between automatic and human translations. In order to improve the descriptive power of current metrics we propose the use of additional syntax-based metrics, and metric combinations inside the QARLA Framework.", "output": {"entities": {"keyphrase": [{"text": "study", "start": 72, "end": 77}, {"text": "Machine Translation Evaluation", "start": 81, "end": 111}, {"text": "criteria", "start": 139, "end": 147}, {"text": "provide", "start": 192, "end": 199}, {"text": "evidence", "start": 210, "end": 218}, {"text": "relationship", "start": 235, "end": 247}, {"text": "kinds", "start": 266, "end": 271}, {"text": "evaluation", "start": 275, "end": 285}, {"text": "point of view", "start": 368, "end": 381}, {"text": "automatic evaluation", "start": 385, "end": 405}, {"text": "metrics", "start": 424, "end": 431}, {"text": "based", "start": 432, "end": 437}, {"text": "system", "start": 478, "end": 484}, {"text": "results", "start": 497, "end": 504}, {"text": "current", "start": 520, "end": 527}, {"text": "evaluation metrics", "start": 528, "end": 546}, {"text": "automatic", "start": 590, "end": 599}, {"text": "translations", "start": 610, "end": 622}, {"text": "order", "start": 627, "end": 632}, {"text": "improve", "start": 636, "end": 643}, {"text": "current", "start": 669, "end": 676}, {"text": "metrics", "start": 677, "end": 684}, {"text": "propose", "start": 688, "end": 695}, {"text": "syntax-based", "start": 718, "end": 730}, {"text": "metrics", "start": 731, "end": 738}, {"text": "metric", "start": 744, "end": 750}, {"text": "combinations", "start": 751, "end": 763}, {"text": "Framework", "start": 781, "end": 790}]}, "relations": {"no_relation": [{"head": {"text": "metrics", "start": 731, "end": 738}, "tail": {"text": "Framework", "start": 781, "end": 790}}]}}, "schema": []} |
| {"input": "Examining The Content Load Of Part Of Speech Blocks For Information Retrieval We investigate the connection between part of speech (POS) distribution and content in language. We define POS blocks to be groups of parts of speech. We hypothesise that there exists a directly proportional relation between the frequency of POS blocks and their content salience. We also hypothesise that the class membership of the parts of speech within such blocks reflects the content load of the blocks, on the basis that open class parts of speech are more content-bearing than closed class parts of speech. We test these hypotheses in the context of Information Retrieval, by syntactically representing queries, and removing from them content-poor blocks, in line with the aforementioned hypotheses. For our first hypothesis, we induce POS distribution information from a corpus, and approximate the probability of occurrence of POS blocks as per two statistical estimators separately. For our second hypothesis, we use simple heuristics to estimate the content load within POS blocks. We use the Text REtrieval Conference (TREC) queries of 1999 and 2000 to retrieve documents from the WT2G and WT10G test collections, with five different retrieval strategies. Experimental outcomes confirm that our hypotheses hold in the context of Information Retrieval.", "output": {"entities": {"keyphrase": [{"text": "part of speech", "start": 116, "end": 130}, {"text": "distribution", "start": 137, "end": 149}, {"text": "content", "start": 154, "end": 161}, {"text": "language", "start": 165, "end": 173}, {"text": "parts of speech", "start": 212, "end": 227}, {"text": "relation", "start": 286, "end": 294}, {"text": "frequency", "start": 307, "end": 316}, {"text": "content", "start": 341, "end": 348}, {"text": "class", "start": 388, "end": 393}, {"text": "parts of speech", "start": 412, "end": 427}, {"text": "content", "start": 460, "end": 467}, {"text": "basis", "start": 495, "end": 500}, {"text": "class", "start": 511, "end": 516}, {"text": "parts of speech", "start": 517, "end": 532}, {"text": "content-bearing", "start": 542, "end": 557}, {"text": "class", "start": 570, "end": 575}, {"text": "parts of speech", "start": 576, "end": 591}, {"text": "test", "start": 596, "end": 600}, {"text": "hypotheses", "start": 607, "end": 617}, {"text": "context", "start": 625, "end": 632}, {"text": "Information Retrieval", "start": 636, "end": 657}, {"text": "queries", "start": 689, "end": 696}, {"text": "content-poor", "start": 721, "end": 733}, {"text": "hypotheses", "start": 774, "end": 784}, {"text": "hypothesis", "start": 800, "end": 810}, {"text": "distribution", "start": 826, "end": 838}, {"text": "information", "start": 839, "end": 850}, {"text": "corpus", "start": 858, "end": 864}, {"text": "probability", "start": 886, "end": 897}, {"text": "occurrence", "start": 901, "end": 911}, {"text": "statistical", "start": 937, "end": 948}, {"text": "hypothesis", "start": 987, "end": 997}, {"text": "simple", "start": 1006, "end": 1012}, {"text": "content", "start": 1040, "end": 1047}, {"text": "Text REtrieval Conference", "start": 1083, "end": 1108}, {"text": "queries", "start": 1116, "end": 1123}, {"text": "documents", "start": 1153, "end": 1162}, {"text": "test", "start": 1187, "end": 1191}, {"text": "collections", "start": 1192, "end": 1203}, {"text": "retrieval", "start": 1225, "end": 1234}, {"text": "strategies", "start": 1235, "end": 1245}, {"text": "Experimental", "start": 1247, "end": 1259}, {"text": "outcomes", "start": 1260, "end": 1268}, {"text": "hypotheses", "start": 1286, "end": 1296}, {"text": "context", "start": 1309, "end": 1316}, {"text": "Information Retrieval", "start": 1320, "end": 1341}]}, "relations": {"no_relation": [{"head": {"text": "information", "start": 839, "end": 850}, "tail": {"text": "corpus", "start": 858, "end": 864}}]}}, "schema": []} |
| {"input": "Integrated Morphological And Syntactic Disambiguation For Modern Hebrew Current parsing models are not immediately applicable for languages that exhibit strong interaction between morphology and syntax, e. g., Modern Hebrew (MH), Arabic and other Semitic languages. This work represents a first attempt at modeling morphological-syntactic interaction in a generative probabilistic framework to allow for MH parsing. We show that morphological information selected in tandem with syntactic categories is instrumental for parsing Semitic languages. We further show that redundant morphological information helps syntactic disambiguation.", "output": {"entities": {"keyphrase": [{"text": "Current", "start": 72, "end": 79}, {"text": "parsing", "start": 80, "end": 87}, {"text": "models", "start": 88, "end": 94}, {"text": "languages", "start": 130, "end": 139}, {"text": "interaction", "start": 160, "end": 171}, {"text": "morphology", "start": 180, "end": 190}, {"text": "syntax", "start": 195, "end": 201}, {"text": "languages", "start": 255, "end": 264}, {"text": "modeling", "start": 306, "end": 314}, {"text": "morphological-syntactic", "start": 315, "end": 338}, {"text": "interaction", "start": 339, "end": 350}, {"text": "framework", "start": 381, "end": 390}, {"text": "parsing", "start": 407, "end": 414}, {"text": "information", "start": 443, "end": 454}, {"text": "syntactic categories", "start": 479, "end": 499}, {"text": "parsing", "start": 520, "end": 527}, {"text": "languages", "start": 536, "end": 545}, {"text": "information", "start": 592, "end": 603}, {"text": "helps", "start": 604, "end": 609}, {"text": "syntactic", "start": 610, "end": 619}, {"text": "disambiguation", "start": 620, "end": 634}]}, "relations": {"no_relation": [{"head": {"text": "models", "start": 88, "end": 94}, "tail": {"text": "languages", "start": 130, "end": 139}}, {"head": {"text": "framework", "start": 381, "end": 390}, "tail": {"text": "parsing", "start": 407, "end": 414}}, {"head": {"text": "parsing", "start": 520, "end": 527}, "tail": {"text": "languages", "start": 536, "end": 545}}, {"head": {"text": "information", "start": 592, "end": 603}, "tail": {"text": "disambiguation", "start": 620, "end": 634}}]}}, "schema": []} |
| {"input": "Lexical Parallelism In Text Structure Determination And Content Analysis In this paper the problem is discussed about the text structure determination and content analysis by lexical parallelism, or the repetition of lexical items. Intersentential relations are determined through the identical, partly identical or lexico-semantic repetition in Japanese scientific texts. Lexical parallelism ratio and lexical parallelism indicator distance are obtained on computer and by hand. And the application of the characteristics to automatic content analysis is dicsussed.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 81, "end": 86}, {"text": "problem", "start": 91, "end": 98}, {"text": "text structure", "start": 122, "end": 136}, {"text": "determination", "start": 137, "end": 150}, {"text": "content", "start": 155, "end": 162}, {"text": "analysis", "start": 163, "end": 171}, {"text": "lexical", "start": 175, "end": 182}, {"text": "repetition", "start": 203, "end": 213}, {"text": "lexical items", "start": 217, "end": 230}, {"text": "relations", "start": 248, "end": 257}, {"text": "lexico-semantic", "start": 316, "end": 331}, {"text": "repetition", "start": 332, "end": 342}, {"text": "Japanese", "start": 346, "end": 354}, {"text": "texts", "start": 366, "end": 371}, {"text": "Lexical", "start": 373, "end": 380}, {"text": "parallelism", "start": 381, "end": 392}, {"text": "ratio", "start": 393, "end": 398}, {"text": "lexical", "start": 403, "end": 410}, {"text": "parallelism", "start": 411, "end": 422}, {"text": "indicator", "start": 423, "end": 432}, {"text": "distance", "start": 433, "end": 441}, {"text": "computer", "start": 458, "end": 466}, {"text": "hand", "start": 474, "end": 478}, {"text": "application", "start": 488, "end": 499}, {"text": "characteristics", "start": 507, "end": 522}, {"text": "automatic", "start": 526, "end": 535}, {"text": "content", "start": 536, "end": 543}, {"text": "analysis", "start": 544, "end": 552}]}, "relations": {"no_relation": [{"head": {"text": "repetition", "start": 203, "end": 213}, "tail": {"text": "lexical items", "start": 217, "end": 230}}, {"head": {"text": "repetition", "start": 332, "end": 342}, "tail": {"text": "texts", "start": 366, "end": 371}}]}}, "schema": []} |
| {"input": "Toward A Parsing Method For Free Word Order Languages \" Free word order \" is a traditional term that should not be taken literally. However, we shall retain the term for its conciseness. Formal descriptionsof syntax have been usually based either on the immediate constituents or on the dependency philosophy. Neither of them seems directly applicable to free word order languages. The intertwining phrases cannot be described naturally by IC rules. Some coordinate constructions are difficult to describe by means of dependency relations. In our opinion, parsers for free word order languages should not be based on the methods developed within the IC framework. Scarce experiments with parsers based on the dependency formalism, eg./5/, do not seem promising. Therefore, we decided to take a fresh start and to attack the problem by reanalys-ing the basic notions of syntax and parsing. We focus our attention on those formal aspects of a language system which might be most useful for automatic text processing. We assume that the morphological level is described along the lines of/2A \"", "output": {"entities": {"keyphrase": [{"text": "word", "start": 61, "end": 65}, {"text": "order", "start": 66, "end": 71}, {"text": "term", "start": 91, "end": 95}, {"text": "term", "start": 161, "end": 165}, {"text": "syntax", "start": 209, "end": 215}, {"text": "based", "start": 234, "end": 239}, {"text": "constituents", "start": 264, "end": 276}, {"text": "dependency", "start": 287, "end": 297}, {"text": "free word order languages", "start": 355, "end": 380}, {"text": "phrases", "start": 399, "end": 406}, {"text": "rules", "start": 443, "end": 448}, {"text": "constructions", "start": 466, "end": 479}, {"text": "dependency relations", "start": 518, "end": 538}, {"text": "opinion", "start": 547, "end": 554}, {"text": "parsers", "start": 556, "end": 563}, {"text": "free word order languages", "start": 568, "end": 593}, {"text": "based", "start": 608, "end": 613}, {"text": "methods", "start": 621, "end": 628}, {"text": "developed", "start": 629, "end": 638}, {"text": "framework", "start": 653, "end": 662}, {"text": "experiments", "start": 671, "end": 682}, {"text": "parsers", "start": 688, "end": 695}, {"text": "based", "start": 696, "end": 701}, {"text": "dependency", "start": 709, "end": 719}, {"text": "formalism", "start": 720, "end": 729}, {"text": "problem", "start": 824, "end": 831}, {"text": "basic", "start": 852, "end": 857}, {"text": "notions", "start": 858, "end": 865}, {"text": "syntax", "start": 869, "end": 875}, {"text": "parsing", "start": 880, "end": 887}, {"text": "focus", "start": 892, "end": 897}, {"text": "aspects", "start": 928, "end": 935}, {"text": "language system", "start": 941, "end": 956}, {"text": "automatic", "start": 988, "end": 997}, {"text": "text processing", "start": 998, "end": 1013}, {"text": "level", "start": 1048, "end": 1053}]}, "relations": {"no_relation": [{"head": {"text": "rules", "start": 443, "end": 448}, "tail": {"text": "phrases", "start": 399, "end": 406}}, {"head": {"text": "methods", "start": 621, "end": 628}, "tail": {"text": "parsers", "start": 556, "end": 563}}, {"head": {"text": "formalism", "start": 720, "end": 729}, "tail": {"text": "parsers", "start": 688, "end": 695}}, {"head": {"text": "aspects", "start": 928, "end": 935}, "tail": {"text": "language system", "start": 941, "end": 956}}]}}, "schema": []} |
| {"input": "Logic Form Transformation Of WordNet And Its Applicability To Question Answering WordNet is a rich source of world knowledge from which formal axioms can be derived. In this paper we present a method for transforming the WordNet glosses into logic forms and further into axioms. The transformation of Word-Net glosses into logic forms is useful for theorem proving and other applications. The paper demonstrates the utility of the WordNet axioms in a question answering system to rank and extract answers.", "output": {"entities": {"keyphrase": [{"text": "source", "start": 99, "end": 105}, {"text": "world knowledge", "start": 109, "end": 124}, {"text": "axioms", "start": 143, "end": 149}, {"text": "paper", "start": 174, "end": 179}, {"text": "method", "start": 193, "end": 199}, {"text": "logic", "start": 242, "end": 247}, {"text": "forms", "start": 248, "end": 253}, {"text": "axioms", "start": 271, "end": 277}, {"text": "transformation", "start": 283, "end": 297}, {"text": "Word-", "start": 301, "end": 306}, {"text": "logic", "start": 323, "end": 328}, {"text": "forms", "start": 329, "end": 334}, {"text": "theorem", "start": 349, "end": 356}, {"text": "applications", "start": 375, "end": 387}, {"text": "paper", "start": 393, "end": 398}, {"text": "utility", "start": 416, "end": 423}, {"text": "axioms", "start": 439, "end": 445}, {"text": "question", "start": 451, "end": 459}, {"text": "system", "start": 470, "end": 476}, {"text": "rank", "start": 480, "end": 484}, {"text": "extract", "start": 489, "end": 496}]}, "relations": {"no_relation": [{"head": {"text": "axioms", "start": 143, "end": 149}, "tail": {"text": "source", "start": 99, "end": 105}}, {"head": {"text": "paper", "start": 174, "end": 179}, "tail": {"text": "method", "start": 193, "end": 199}}, {"head": {"text": "axioms", "start": 439, "end": 445}, "tail": {"text": "system", "start": 470, "end": 476}}]}}, "schema": []} |
| {"input": "Parsing Non-Recursive CFGs We consider the problem of parsing non-recursive context-free grammars, i. e., context-free grammars that generate finite languages. In natural language processing, this problem arises in several areas of application, including natural language generation, speech recognition and machine translation. We present two tabular algorithms for parsing ofnon-recursive context-free grammars, and show that they perform well in practical settings, despite the fact that this problem is PSPACE-complete.", "output": {"entities": {"keyphrase": [{"text": "problem", "start": 43, "end": 50}, {"text": "parsing", "start": 54, "end": 61}, {"text": "context-free", "start": 76, "end": 88}, {"text": "context-free", "start": 106, "end": 118}, {"text": "generate", "start": 133, "end": 141}, {"text": "languages", "start": 149, "end": 158}, {"text": "natural language processing", "start": 163, "end": 190}, {"text": "problem", "start": 197, "end": 204}, {"text": "areas", "start": 223, "end": 228}, {"text": "application", "start": 232, "end": 243}, {"text": "including", "start": 245, "end": 254}, {"text": "natural language generation", "start": 255, "end": 282}, {"text": "speech recognition", "start": 284, "end": 302}, {"text": "machine translation", "start": 307, "end": 326}, {"text": "algorithms", "start": 351, "end": 361}, {"text": "parsing", "start": 366, "end": 373}, {"text": "context-free", "start": 390, "end": 402}, {"text": "perform", "start": 432, "end": 439}, {"text": "problem", "start": 495, "end": 502}]}, "relations": {"no_relation": [{"head": {"text": "algorithms", "start": 351, "end": 361}, "tail": {"text": "parsing", "start": 366, "end": 373}}]}}, "schema": []} |
| {"input": "Massive Disambiguation Of Large Text Corpora With Flexible Categorial Grammar A new method of automatic lexical disambiguation of big texts is described, using recent proof-theoretical results from the theory of categorial grammar.", "output": {"entities": {"keyphrase": [{"text": "method", "start": 84, "end": 90}, {"text": "automatic", "start": 94, "end": 103}, {"text": "lexical", "start": 104, "end": 111}, {"text": "disambiguation", "start": 112, "end": 126}, {"text": "texts", "start": 134, "end": 139}, {"text": "results", "start": 185, "end": 192}, {"text": "theory", "start": 202, "end": 208}, {"text": "categorial grammar", "start": 212, "end": 230}]}, "relations": {}}, "schema": []} |
| {"input": "Using Constraints In A Constructive Version Of GPSG Complex categories are caracteristic of unification grammars as for example GPSG [Shieber86a]. They are sets of pairs of features and values. The unification, which can be applied to two or more categories, is the essential operation. The papers of [Shieber85], [Barton85] and [Ristad86] deal with the influence of complex categories on the efficiency of the parsing algorithm. This is one problem from using complex categories, another one arises when using a constructive version of GPSG (see [Busemann/Hauenschild88] in this volume). Namely that the application of admissibility conditions, e. g. LP statements and FCRs to m a local tree t is prevented because particular feature values of categories in t are not yet specified, but they will be instantiated later somewhere else in the complete tree. Similar problems are described. in [Karttunen86] for D-PATR. This work describes the latter problem and will present a solution working with computation, evaluation and propagation of constraints within local trees (depth 1). The constraint evaluation will reject local trees if the constraints of the subtrees of the daughters are violated.", "output": {"entities": {"keyphrase": [{"text": "Complex", "start": 52, "end": 59}, {"text": "categories", "start": 60, "end": 70}, {"text": "unification", "start": 92, "end": 103}, {"text": "example", "start": 120, "end": 127}, {"text": "pairs", "start": 164, "end": 169}, {"text": "features", "start": 173, "end": 181}, {"text": "unification", "start": 198, "end": 209}, {"text": "applied", "start": 224, "end": 231}, {"text": "categories", "start": 247, "end": 257}, {"text": "operation", "start": 276, "end": 285}, {"text": "papers", "start": 291, "end": 297}, {"text": "deal", "start": 340, "end": 344}, {"text": "influence", "start": 354, "end": 363}, {"text": "complex", "start": 367, "end": 374}, {"text": "categories", "start": 375, "end": 385}, {"text": "efficiency", "start": 393, "end": 403}, {"text": "parsing", "start": 411, "end": 418}, {"text": "algorithm", "start": 419, "end": 428}, {"text": "problem", "start": 442, "end": 449}, {"text": "complex", "start": 461, "end": 468}, {"text": "categories", "start": 469, "end": 479}, {"text": "version", "start": 526, "end": 533}, {"text": "volume", "start": 580, "end": 586}, {"text": "application", "start": 605, "end": 616}, {"text": "tree", "start": 688, "end": 692}, {"text": "feature values", "start": 727, "end": 741}, {"text": "categories", "start": 745, "end": 755}, {"text": "tree", "start": 851, "end": 855}, {"text": "problems", "start": 865, "end": 873}, {"text": "problem", "start": 949, "end": 956}, {"text": "solution", "start": 976, "end": 984}, {"text": "computation", "start": 998, "end": 1009}, {"text": "evaluation", "start": 1011, "end": 1021}, {"text": "propagation", "start": 1026, "end": 1037}, {"text": "constraints", "start": 1041, "end": 1052}, {"text": "trees", "start": 1066, "end": 1071}, {"text": "constraint", "start": 1087, "end": 1097}, {"text": "evaluation", "start": 1098, "end": 1108}, {"text": "trees", "start": 1127, "end": 1132}, {"text": "constraints", "start": 1140, "end": 1151}]}, "relations": {"no_relation": [{"head": {"text": "feature values", "start": 727, "end": 741}, "tail": {"text": "categories", "start": 745, "end": 755}}]}}, "schema": []} |
| {"input": "Phonological Processing Of Speech Variants This paper describes a strategy for the extension of the phonological lexicon in order that nonstandard forms which arise in fast speech may be processed by a speech recognition system. By way of illustration, an outline of the phonological processing of standard wordforms by the phonological parser (PhoPa) is given and then the extension procedure which is based on this phonological parser is discussed. The lexicon extension procedure has two stages: phonotactic extension which involves the introduction of additional restrictions into the phonotactic network for the standard language in the form of metarules describing phonological processes, and specialised word model construction whereby for each standard phonemic wordform a verification net which contains all variants of this standard form is compiled. The complete system serves as a phonologically oriented lexicon development tool, and its theoretical interest lies in its contribution to the field of speech variant learning.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 48, "end": 53}, {"text": "strategy", "start": 66, "end": 74}, {"text": "extension", "start": 83, "end": 92}, {"text": "lexicon", "start": 113, "end": 120}, {"text": "order", "start": 124, "end": 129}, {"text": "forms", "start": 147, "end": 152}, {"text": "speech", "start": 173, "end": 179}, {"text": "processed", "start": 187, "end": 196}, {"text": "speech recognition system", "start": 202, "end": 227}, {"text": "outline", "start": 256, "end": 263}, {"text": "processing", "start": 284, "end": 294}, {"text": "standard", "start": 298, "end": 306}, {"text": "parser", "start": 337, "end": 343}, {"text": "extension", "start": 374, "end": 383}, {"text": "procedure", "start": 384, "end": 393}, {"text": "based", "start": 403, "end": 408}, {"text": "parser", "start": 430, "end": 436}, {"text": "lexicon", "start": 455, "end": 462}, {"text": "extension", "start": 463, "end": 472}, {"text": "procedure", "start": 473, "end": 482}, {"text": "extension", "start": 511, "end": 520}, {"text": "introduction", "start": 540, "end": 552}, {"text": "restrictions", "start": 567, "end": 579}, {"text": "network", "start": 601, "end": 608}, {"text": "standard", "start": 617, "end": 625}, {"text": "language", "start": 626, "end": 634}, {"text": "form", "start": 642, "end": 646}, {"text": "processes", "start": 684, "end": 693}, {"text": "word model", "start": 711, "end": 721}, {"text": "construction", "start": 722, "end": 734}, {"text": "standard", "start": 752, "end": 760}, {"text": "verification", "start": 781, "end": 793}, {"text": "variants", "start": 817, "end": 825}, {"text": "standard", "start": 834, "end": 842}, {"text": "form", "start": 843, "end": 847}, {"text": "system", "start": 874, "end": 880}, {"text": "lexicon", "start": 917, "end": 924}, {"text": "development", "start": 925, "end": 936}, {"text": "tool", "start": 937, "end": 941}, {"text": "contribution", "start": 984, "end": 996}, {"text": "field", "start": 1004, "end": 1009}, {"text": "speech", "start": 1013, "end": 1019}, {"text": "variant", "start": 1020, "end": 1027}, {"text": "learning", "start": 1028, "end": 1036}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 48, "end": 53}, "tail": {"text": "strategy", "start": 66, "end": 74}}, {"head": {"text": "speech recognition system", "start": 202, "end": 227}, "tail": {"text": "speech", "start": 173, "end": 179}}, {"head": {"text": "parser", "start": 337, "end": 343}, "tail": {"text": "processing", "start": 284, "end": 294}}]}}, "schema": []} |
| {"input": "Zero Pronouns As Experiencer In Japanese Discourse The process of finding the antecedent of zero pronoun, that is indispensable to Japanese language understanding, is the topic of this paper. Here we mainly concern with discourses comprising two sentences that are in a subordinate relation, especially one of them describes the agent' s volitional action and the other describes the reason of the action. We propose basically two new principles: (1) The agent of an action should experience a certain psychological reason, (2) Predicates reporting someone' s psychological state are categorized into 1) weakly or 2) strongly bound to the expected point of view. Combination of these principles accounts for some problematic Japanese zero anaphora, which cannot be accounted for by the theories so far proposed.", "output": {"entities": {"keyphrase": [{"text": "process", "start": 55, "end": 62}, {"text": "Japanese", "start": 131, "end": 139}, {"text": "language understanding", "start": 140, "end": 162}, {"text": "topic", "start": 171, "end": 176}, {"text": "paper", "start": 185, "end": 190}, {"text": "concern", "start": 207, "end": 214}, {"text": "discourses", "start": 220, "end": 230}, {"text": "sentences", "start": 246, "end": 255}, {"text": "relation", "start": 282, "end": 290}, {"text": "agent", "start": 329, "end": 334}, {"text": "action", "start": 349, "end": 355}, {"text": "reason", "start": 384, "end": 390}, {"text": "action", "start": 398, "end": 404}, {"text": "propose", "start": 409, "end": 416}, {"text": "principles", "start": 435, "end": 445}, {"text": "agent", "start": 455, "end": 460}, {"text": "action", "start": 467, "end": 473}, {"text": "experience", "start": 481, "end": 491}, {"text": "reason", "start": 516, "end": 522}, {"text": "reporting", "start": 539, "end": 548}, {"text": "point of view", "start": 648, "end": 661}, {"text": "Combination", "start": 663, "end": 674}, {"text": "principles", "start": 684, "end": 694}, {"text": "Japanese", "start": 725, "end": 733}, {"text": "theories", "start": 786, "end": 794}, {"text": "proposed", "start": 802, "end": 810}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 185, "end": 190}, "tail": {"text": "language understanding", "start": 140, "end": 162}}, {"head": {"text": "sentences", "start": 246, "end": 255}, "tail": {"text": "discourses", "start": 220, "end": 230}}]}}, "schema": []} |
| {"input": "Tokenization As The Initial Phase In NLP In this paper, the authors address the significance and complexity of tokenization, the beginning step of NLP. Notions of word and token arc discussed and defined from the viewpoints of lexicography and pragmatic implementation, respectively. Automatic segmentation of Chinese words is presented as an illustration of tokenization. Practical approaches to identification of compound tokens in English, such as idioms, phrasal verbs and fixed expressions, are developed.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 49, "end": 54}, {"text": "significance", "start": 80, "end": 92}, {"text": "complexity", "start": 97, "end": 107}, {"text": "step", "start": 139, "end": 143}, {"text": "word", "start": 163, "end": 167}, {"text": "arc", "start": 178, "end": 181}, {"text": "implementation", "start": 254, "end": 268}, {"text": "Automatic", "start": 284, "end": 293}, {"text": "Chinese", "start": 310, "end": 317}, {"text": "words", "start": 318, "end": 323}, {"text": "approaches", "start": 383, "end": 393}, {"text": "identification", "start": 397, "end": 411}, {"text": "English", "start": 434, "end": 441}, {"text": "idioms", "start": 451, "end": 457}, {"text": "verbs", "start": 467, "end": 472}, {"text": "expressions", "start": 483, "end": 494}, {"text": "developed", "start": 500, "end": 509}]}, "relations": {}}, "schema": []} |
| {"input": "An English-To-Korean Machine Translator: MATES/EK This note introduces an EnglishKorean Machine Translation System MATES/EK, which has been de veloped as a research prototype and still under upgrading KAIST (Korea Advanced Institute of Science and Technology). MATES/EK a transfer-based system and has several subsystems that can be used support other-developments. They are grammar developing environment systems, dictionary developing tools, a set augmented context free grain mars English syntactic analysis, and so on.", "output": {"entities": {"keyphrase": [{"text": "note", "start": 55, "end": 59}, {"text": "Machine Translation System", "start": 88, "end": 114}, {"text": "research", "start": 156, "end": 164}, {"text": "prototype", "start": 165, "end": 174}, {"text": "Science", "start": 236, "end": 243}, {"text": "Technology", "start": 248, "end": 258}, {"text": "transfer-based system", "start": 272, "end": 293}, {"text": "support", "start": 338, "end": 345}, {"text": "developments", "start": 352, "end": 364}, {"text": "developing", "start": 383, "end": 393}, {"text": "environment", "start": 394, "end": 405}, {"text": "systems", "start": 406, "end": 413}, {"text": "dictionary", "start": 415, "end": 425}, {"text": "tools", "start": 437, "end": 442}, {"text": "context", "start": 460, "end": 467}, {"text": "mars", "start": 479, "end": 483}, {"text": "English", "start": 484, "end": 491}, {"text": "syntactic analysis", "start": 492, "end": 510}]}, "relations": {}}, "schema": []} |
| {"input": "Notes On LR Parser Design This paper discusses the design of an LR parser for a specific high-coverage English grammar. The design principles, though, are applicable to a large class of unification-based grammars where the constraints are realized as Prolog terms and applied monotonically through instantiation, where there is no right movement, and where left movement is handled by gap threading. The LR parser was constructed for experiments on probabilistic parsing and speedup learning, see [10]. LR parsers are suitable for probabilistic parsing since they contain a representation of the current parsing state, namely the stack and the input string, and since the actions of the parsing tables are easily attributed probabilities conditional on this parsing state. LR parsers are suitable for the speedup learning application since the learneu grammar is much larger than the original grammar, and the prefixes of the learned rules overlap to a very high degree, circumstances that are far from ideal for the system' s original parser. Even though these ends influenced the design of the parser, this article does not focus on these applications but rather on the design and testing of the parser itself.outputentitieskeyphrasetextpaperstartendtextdesignstartendtextparserstartendtexthigh-coveragestartendtextEnglishstartendtextdesignstartendtextprinciplesstartendtextclassstartendtextunification-basedstartendtextconstraintsstartendtexttermsstartendtextappliedstartendtextgapstartendtextthreadingstartendtextparserstartendtextconstructedstartendtextexperimentsstartendtextparsingstartendtextlearningstartendtextparsersstartendtextparsingstartendtextrepresentationstartendtextcurrentstartendtextparsingstartendtextinputstartendtextstringstartendtextactionsstartendtextparsingstartendtexttablesstartendtextprobabilitiesstartendtextparsingstartendtextparsersstartendtextapplicationstartendtextprefixesstartendtextrulesstartendtextdegreestartendtextsystemstartendtextparserstartendtextinfluencedstartendtextdesignstartendtextparserstartendtextfocusstartendtextapplicationsstartendtextdesignstartendtextparserstartendrelationsno_relationheadtextpaperstartendtailtextparserstartendheadtextparsersstartendtailtextparsingstartendheadtextparsersstartendtailtextapplicationstartendschema |
| inputTowards Discourse-Oriented Nonmonotonic System The purpose of this paper is to analyse the phenomenon of nonmonotonicity in a natural language and to formulate a number of general principles which should be taken into consideration while constructing a discourse oriented nonmonotonic formalism.outputentitieskeyphrasetextpurposestartendtextpaperstartendtextphenomenonstartendtextnatural languagestartendtextnumberstartendtextprinciplesstartendtextconsiderationstartendtextconstructingstartendtextdiscoursestartendtextformalismstartendrelationsno_relationheadtextpaperstartendtailtextformalismstartendschema |
| inputMorphology And Cross Dependencies In The Synthesis Of Personal Pronouns In Romance Languages \ cross dependency \outputentitieskeyphrasetextpaperstartendtextproblemsstartendtextsynthesisstartendtextsystemstartendtextgeneratesstartendtexttextsstartendtextlanguagesstartendtextlevelstartendtextgeneration processstartendtextcrossstartendtextdependencystartendtextphenomenastartendtextsynthesisstartendtextsynthesisstartendtextexamplesstartendtextlanguagesstartendtextrobuststartendtextgeneration systemstartendtextimplementedstartendrelationsno_relationheadtextexamplesstartendtailtextlanguagesstartendschema |
| inputA News Analysis System This paper describes a prototype news analysis system which classifies and indexes news stories in real time. The system extracts stories from newswire, parses the sentences of the story, and then maps the syntactic structures into concept base. This process results in an index containing both general categories and specific details. Central to this system is a Government-Binding parser which processes each sentence of a news item. The system is completely modular and can be interfaced with different news feeds or concept bases.outputentitieskeyphrasetextpaperstartendtextprototypestartendtextnewsstartendtextanalysisstartendtextsystemstartendtextindexesstartendtextnewsstartendtextreal timestartendtextsystemstartendtextextractsstartendtextparsesstartendtextsentencesstartendtextmapsstartendtextsyntactic structuresstartendtextconceptstartendtextbasestartendtextprocessstartendtextresultsstartendtextindexstartendtextcategoriesstartendtextdetailsstartendtextsystemstartendtextparserstartendtextprocessesstartendtextsentencestartendtextnewsstartendtextitemstartendtextsystemstartendtextnewsstartendtextconceptstartendtextbasesstartendrelationsno_relationheadtextpaperstartendtailtextsystemstartendheadtextbasestartendtailtextsyntactic structuresstartendheadtextcategoriesstartendtailtextindexstartendheadtextparserstartendtailtextsystemstartendheadtextsentencestartendtailtextitemstartendschema |
| inputError Diagnosing And Selection In A Training System For Second Language Learning A diagnosing procedure to be used in intelligent systems for language instruction is presented. Based on a knowledge representation scheme for a certain class of syntactic correctness conditions the system carries out a thorough analysis of possible error hypotheses and their consequences. A comparison with earlier attempts shows a clearly improved precision of diagnostic results. First of all, the procedure concentrates on an exact localization of rule violations, but-if desired-is able to infer information about factual faults as well.outputentitieskeyphrasetextprocedurestartendtextsystemsstartendtextlanguagestartendtextinstructionstartendtextBasedstartendtextknowledge representationstartendtextschemestartendtextclassstartendtextsyntacticstartendtextcorrectnessstartendtextsystemstartendtextanalysisstartendtexterrorstartendtexthypothesesstartendtextcomparisonstartendtextimprovedstartendtextprecisionstartendtextresultsstartendtextprocedurestartendtextlocalizationstartendtextrulestartendtextinformationstartendrelationsno_relationheadtextprocedurestartendtailtextsystemsstartendschema |
| inputTapping Huge Temporally Indexed Textual Resources with WCTAnalyze WCTAnalyze is a tool for storing, accessing and visually analyzing huge collections of temporally indexeddata. It is motivated by applications in media analysis, business intelligence etc. where higher level analysis is performed on top of linguistically and statistically processed unstructured textualdata. WCTAnalyze combines fast access with economically storage behaviour and appropriates a lot of built in visualization options for result presentation in detail as well as in contrast. So it enables an efficient and effective way to explore chronological text patterns of word forms, their co-occurrence sets and co-occurrence set intersections. Digging deep into co-occurrences of the same semantic or syntactic describing wordforms, some entities can be recognized as to be temporal related, whereas other differ significantly. This behaviour motivates approaches in interactive discovering events based on co-occurrence subsets.outputentitieskeyphrasetexttoolstartendtextaccessingstartendtextcollectionsstartendtextindexeddatastartendtextindexeddatastartendtextapplicationsstartendtextanalysisstartendtextlevelstartendtextanalysisstartendtextperformedstartendtextprocessedstartendtexttextualdatastartendtextaccessstartendtextstoragestartendtextbehaviourstartendtextvisualizationstartendtextoptionsstartendtextresultstartendtextpresentationstartendtextdetailstartendtextcontraststartendtexttextstartendtextpatternsstartendtextwordstartendtextformsstartendtextco-occurrencestartendtextco-occurrencestartendtextintersectionsstartendtextsemanticstartendtextsyntacticstartendtextentitiesstartendtextbehaviourstartendtextapproachesstartendtexteventsstartendtextbasedstartendtextco-occurrencestartendrelationsno_relationheadtextindexeddatastartendtailtextcollectionsstartendheadtextpatternsstartendtailtextformsstartendschema |
| inputLinguistic Structure and Bilingual Informants Help Induce Machine Translation of Lesser-Resourced Languages Producing machine translation (MT) for the many minority languages in the world is a serious challenge. Minority languages typically have few resources for building MT systems. For many minor languages there is little machine readable text, few knowledgeable linguists, and little money available for MT development. For these reasons, our research programs on minority language MT have focused on leveraging to the maximum extent two resources that are available for minority languages: linguistic structure and bilingual informants. All natural languages contain linguistic structure. And although the details of that linguistic structure vary from language to language, language universals such as context-free syntactic structure and the paradigmatic structure of inflectional morphology, allow us to learn the specific details of a minority language. Similarly, most minority languages possess speakers who are bilingual with the major language of the area. This paper discusses our efforts to utilize linguistic structure and the translation information that bilingual informants can provide in three sub-areas of our rapid development MT program: morphology induction, syntactic transfer rule learning, and refinement of imperfect learned rules.outputentitieskeyphrasetextmachine translationstartendtextlanguagesstartendtextchallengestartendtextlanguagesstartendtextresourcesstartendtextbuildingstartendtextMT systemsstartendtextlanguagesstartendtextmachinestartendtexttextstartendtextlinguistsstartendtextdevelopmentstartendtextreasonsstartendtextresearchstartendtextprogramsstartendtextlanguagestartendtextfocusedstartendtextleveragingstartendtextextentstartendtextresourcesstartendtextlanguagesstartendtextlinguistic structurestartendtextnatural languagesstartendtextlinguistic structurestartendtextdetailsstartendtextlinguistic structurestartendtextlanguagestartendtextlanguagestartendtextlanguagestartendtextcontext-freestartendtextsyntactic structurestartendtextstructurestartendtextmorphologystartendtextdetailsstartendtextlanguagestartendtextlanguagesstartendtextlanguagestartendtextareastartendtextpaperstartendtexteffortsstartendtextlinguistic structurestartendtexttranslationstartendtextinformationstartendtextprovidestartendtextdevelopmentstartendtextprogramstartendtextmorphologystartendtextinductionstartendtextsyntacticstartendtexttransferstartendtextrulestartendtextlearningstartendtextrefinementstartendtextrulesstartendrelationsno_relationheadtextmachine translationstartendtailtextlanguagesstartendheadtextlinguistic structurestartendtailtextnatural languagesstartendheadtextpaperstartendtailtextlinguistic structurestartendschema |
| inputA Generative Model for Parsing Natural Language to Meaning Representations In this paper, we present an algorithm for learning a generative model of natural language sentences together with their formal meaning representations with hierarchical structures. The model is applied to the task of mapping sentences to hierarchical representations of their underlying meaning. We introduce dynamic programming techniques for efficient training and decoding. In experiments, we demonstrate that the model, when coupled with a discriminative reranking technique, achieves state-of-the-art performance when tested on two publicly available corpora. The generative model degrades robustly when presented with instances that are different from those seen in training. This allows a notable improvement in recall compared to previous models.outputentitieskeyphrasetextpaperstartendtextalgorithmstartendtextgenerative modelstartendtextnatural language sentencesstartendtextmeaning representationsstartendtexthierarchical structuresstartendtextmodelstartendtextappliedstartendtexttaskstartendtextmappingstartendtextsentencesstartendtextrepresentationsstartendtextdynamic programmingstartendtexttechniquesstartendtexttrainingstartendtextexperimentsstartendtextmodelstartendtexttechniquestartendtextperformancestartendtexttestedstartendtextcorporastartendtextgenerative modelstartendtextinstancesstartendtexttrainingstartendtextimprovementstartendtextrecallstartendtextmodelsstartendrelationsno_relationheadtextpaperstartendtailtextalgorithmstartendheadtextmeaning representationsstartendtailtextnatural language sentencesstartendheadtextmodelstartendtailtextmappingstartendheadtextmodelstartendtailtextperformancestartendschema |
| inputA Comparative Study on Language Identification Methods In this paper we present two experiments conducted for comparison of different language identification algorithms. Short words-, frequent words-and n-gram-based approaches are considered and combined with the Ad-Hoc Ranking classification method. The language identification process can be subdivided into two main steps: First a document model is generated for the document and a language model for the language; second the language of the document is determined on the basis of the language model and is added to the document as additional information. In this work we present our evaluation results and discuss the importance of a dynamic value for the out-of-place measure.outputentitieskeyphrasetextpaperstartendtextexperimentsstartendtextcomparisonstartendtextlanguagestartendtextidentificationstartendtextalgorithmsstartendtextwords-startendtextwords-startendtextn-gram-based approachesstartendtextclassificationstartendtextmethodstartendtextlanguagestartendtextidentificationstartendtextprocessstartendtextmainstartendtextstepsstartendtextdocumentstartendtextmodelstartendtextgeneratedstartendtextdocumentstartendtextlanguage modelstartendtextlanguagestartendtextlanguagestartendtextdocumentstartendtextbasisstartendtextlanguage modelstartendtextdocumentstartendtextinformationstartendtextevaluation resultsstartendtextimportancestartendrelationsno_relationheadtextpaperstartendtailtextexperimentsstartendheadtextmodelstartendtailtextdocumentstartendheadtextlanguage modelstartendtailtextlanguagestartendschema |
| inputQUALIFIER: Question Answering By Lexical Fabric And External Resources One of the major challenges in TREC-style question-answering (QA) is to overcome the mismatch in the lexical representations in the query space and document space. This is particularly severe in QA as exact answers, rather than documents, are required in response to questions. Most current approaches overcome the mismatch problem by employing eitherdata redundancy strategy through the use of Web or linguistic resources. This paper investigates the integration of lexical relations and Web knowledge to tackle this problem. The results obtained on TREC11 QA corpus indicate that our approach is both feasible and effective.outputentitieskeyphrasetextchallengesstartendtextquestion-answeringstartendtextmismatchstartendtextlexicalstartendtextrepresentationsstartendtextquerystartendtextspacestartendtextdocumentstartendtextspacestartendtextdocumentsstartendtextrequiredstartendtextquestionsstartendtextcurrentstartendtextapproachesstartendtextmismatchstartendtextproblemstartendtexteitherdatastartendtextredundancystartendtextstrategystartendtextlinguistic resourcesstartendtextpaperstartendtextintegrationstartendtextlexicalstartendtextrelationsstartendtextknowledgestartendtextproblemstartendtextresultsstartendtextcorpusstartendtextapproachstartendrelationsno_relationheadtextmismatchstartendtailtextrepresentationsstartendschema |
| inputCALL: The Potential Of LINGWARE And The Use Of Empirical Linguistic Data Language technology has significantly evolved during the last decade. However, the community of language learning seems to ignore this development, most of the existing language learning systems drawing their enhancements from other sources, such as hypertext, multimedia, interactive video, information retrieval. Despite some spectacular progress made at the level of interface, several fundamental language learning principles, are only partially met. Nevertheless, the hypermedia technology did solve one very important aspect of computer-assisted learning by putting the student in a visual environment. Minimizing cultural differences they' ve been able to draw on shared background knowledge (microworld immersiveness). Other important aspects of typical immersion-based approaches, i. e. natural learning, such as mixed-initiative, fault-tolerance, dialogue repair, cooperative behaviour, etc. are still in their infancy. In real settings learners freely interact with their environment (parents, tutors), taking turns, asking for explanations, shifting topics, etc. The language produced by the learner is more often than not agrammatical, yet this does not prevent the tutor to proceed with the dialog. Error correction is usually done contextually, by drawing either explicitly attention to the deviation, by producing a similar but correct sentence, or by simply ignoring the mistake leaving its correction for later. There are many AI and CL programs solving various specific CALL-relevant problems. If assembled properly, these pieces could result in very powerful language learning systems. Lexical thesauri Parsers Generators Semantic interpreters/generators", "output": {"entities": {"keyphrase": [{"text": "Language technology", "start": 73, "end": 92}, {"text": "community", "start": 156, "end": 165}, {"text": "language", "start": 169, "end": 177}, {"text": "development", "start": 208, "end": 219}, {"text": "language", "start": 242, "end": 250}, {"text": "systems", "start": 260, "end": 267}, {"text": "enhancements", "start": 282, "end": 294}, {"text": "sources", "start": 306, "end": 313}, {"text": "hypertext", "start": 323, "end": 332}, {"text": "video", "start": 358, "end": 363}, {"text": "information retrieval", "start": 365, "end": 386}, {"text": "progress", "start": 413, "end": 421}, {"text": "level", "start": 434, "end": 439}, {"text": "interface", "start": 443, "end": 452}, {"text": "language", "start": 474, "end": 482}, {"text": "principles", "start": 492, "end": 502}, {"text": "technology", "start": 557, "end": 567}, {"text": "solve", "start": 572, "end": 577}, {"text": "aspect", "start": 597, "end": 603}, {"text": "computer-assisted", "start": 607, "end": 624}, {"text": "learning", "start": 625, "end": 633}, {"text": "environment", "start": 669, "end": 680}, {"text": "differences", "start": 702, "end": 713}, {"text": "background knowledge", "start": 751, "end": 771}, {"text": "aspects", "start": 816, "end": 823}, {"text": "immersion-based approaches", "start": 835, "end": 861}, {"text": "natural", "start": 869, "end": 876}, {"text": "learning", "start": 877, "end": 885}, {"text": "mixed-initiative", "start": 895, "end": 911}, {"text": "dialogue", "start": 930, "end": 938}, {"text": "repair", "start": 939, "end": 945}, {"text": "behaviour", "start": 959, "end": 968}, {"text": "environment", "start": 1056, "end": 1067}, {"text": "explanations", "start": 1112, "end": 1124}, {"text": "shifting", "start": 1126, "end": 1134}, {"text": "topics", "start": 1135, "end": 1141}, {"text": "language", "start": 1152, "end": 1160}, {"text": "dialog", "start": 1278, "end": 1284}, {"text": "Error", "start": 1286, "end": 1291}, {"text": "correction", "start": 1292, "end": 1302}, {"text": "deviation", "start": 1379, "end": 1388}, {"text": "sentence", "start": 1425, "end": 1433}, {"text": "correction", "start": 1481, "end": 1491}, {"text": "programs", "start": 1528, "end": 1536}, {"text": "solving", "start": 1537, "end": 1544}, {"text": "CALL-relevant", "start": 1562, "end": 1575}, {"text": "problems", "start": 1576, "end": 1584}, {"text": "result", "start": 1628, "end": 1634}, {"text": "language", "start": 1652, "end": 1660}, {"text": "systems", "start": 1670, "end": 1677}, {"text": "Lexical", "start": 1679, "end": 1686}, {"text": "thesauri", "start": 1687, "end": 1695}, {"text": "Semantic", "start": 1715, "end": 1723}, {"text": "interpreters", "start": 1724, "end": 1736}, {"text": "generators", "start": 1737, "end": 1747}]}, "relations": {}}, "schema": []} |
| {"input": "Named Entity Chunking Techniques In Supervised Learning For Japanese Named Entity Recognition This paper focuses on the issue of named entity chunking in Japanese named entity recognition. We apply the supervised decision list learning method to Japanese named entity recognition. We also investigate and incorporate several named-entity noun phrase chunking techniques and experimentally evaluate and compare their performance. In addition, we propose a method for incorporating richer contextual information as well as patterns of constituent morphemes within a named entity, which have not been considered in previous research, and show that the proposed method outperforms these previous approaches.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 99, "end": 104}, {"text": "focuses", "start": 105, "end": 112}, {"text": "issue", "start": 120, "end": 125}, {"text": "named", "start": 129, "end": 134}, {"text": "entity", "start": 135, "end": 141}, {"text": "chunking", "start": 142, "end": 150}, {"text": "Japanese", "start": 154, "end": 162}, {"text": "named", "start": 163, "end": 168}, {"text": "entity", "start": 169, "end": 175}, {"text": "recognition", "start": 176, "end": 187}, {"text": "apply", "start": 192, "end": 197}, {"text": "decision", "start": 213, "end": 221}, {"text": "list", "start": 222, "end": 226}, {"text": "method", "start": 236, "end": 242}, {"text": "Japanese", "start": 246, "end": 254}, {"text": "named", "start": 255, "end": 260}, {"text": "entity", "start": 261, "end": 267}, {"text": "recognition", "start": 268, "end": 279}, {"text": "named-entity", "start": 325, "end": 337}, {"text": "noun phrase", "start": 338, "end": 349}, {"text": "chunking", "start": 350, "end": 358}, {"text": "techniques", "start": 359, "end": 369}, {"text": "evaluate", "start": 389, "end": 397}, {"text": "performance", "start": 416, "end": 427}, {"text": "addition", "start": 432, "end": 440}, {"text": "propose", "start": 445, "end": 452}, {"text": "method", "start": 455, "end": 461}, {"text": "contextual information", "start": 487, "end": 509}, {"text": "patterns", "start": 521, "end": 529}, {"text": "constituent", "start": 533, "end": 544}, {"text": "named", "start": 564, "end": 569}, {"text": "entity", "start": 570, "end": 576}, {"text": "research", "start": 621, "end": 629}, {"text": "proposed", "start": 649, "end": 657}, {"text": "method", "start": 658, "end": 664}, {"text": "approaches", "start": 692, "end": 702}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 99, "end": 104}, "tail": {"text": "chunking", "start": 142, "end": 150}}, {"head": {"text": "method", "start": 236, "end": 242}, "tail": {"text": "recognition", "start": 268, "end": 279}}]}}, "schema": []} |
| {"input": "An Inference-Based Approach To Dialogue System Design We present an architecture for spoken dialogue systems where first-order inference (both theorem proving and model building) plays a crucial role in interpreting utterances of dialogue participants and deciding how the system should respond and carry out instructions. The dialogue itself is represented SIS St DRS which is translated into first-order logic for inference tasks. The system is implemented SIS St society of OAA-agents, and evaluated against a specific application (home automation).", "output": {"entities": {"keyphrase": [{"text": "architecture", "start": 68, "end": 80}, {"text": "dialogue systems", "start": 92, "end": 108}, {"text": "first-order", "start": 115, "end": 126}, {"text": "inference", "start": 127, "end": 136}, {"text": "theorem", "start": 143, "end": 150}, {"text": "model", "start": 163, "end": 168}, {"text": "building", "start": 169, "end": 177}, {"text": "role", "start": 195, "end": 199}, {"text": "utterances", "start": 216, "end": 226}, {"text": "dialogue", "start": 230, "end": 238}, {"text": "participants", "start": 239, "end": 251}, {"text": "system", "start": 273, "end": 279}, {"text": "instructions", "start": 309, "end": 321}, {"text": "dialogue", "start": 327, "end": 335}, {"text": "SIS", "start": 358, "end": 361}, {"text": "translated", "start": 378, "end": 388}, {"text": "first-order", "start": 394, "end": 405}, {"text": "logic", "start": 406, "end": 411}, {"text": "inference", "start": 416, "end": 425}, {"text": "tasks", "start": 426, "end": 431}, {"text": "system", "start": 437, "end": 443}, {"text": "implemented", "start": 447, "end": 458}, {"text": "SIS", "start": 459, "end": 462}, {"text": "evaluated", "start": 493, "end": 502}, {"text": "application", "start": 522, "end": 533}, {"text": "automation", "start": 540, "end": 550}]}, "relations": {}}, "schema": []} |
| {"input": "Compilation Of Unification Grammars With Compositional Semantics To Speech Recognition Packages In this paper a method to compile unification grammars into speech recognition packages is presented, and in particular, rules are specified to transfer the compositional semantics stated in unification grammars into speech recognition grammars. The resulting compiler creates a context-free backbone of the unification grammar, eliminates left-recursive productions and removes redundant grammar rules. The method was tested on a medium-sized unification grammar for English using Nuance speech recognition software on a corpus of 131 utterances of 12 different speakers. Results showed no significant computational overhead with respect to speech recognition performances for speech recognition grammar with compositional semantics compared to grammars without.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 104, "end": 109}, {"text": "method", "start": 112, "end": 118}, {"text": "unification", "start": 130, "end": 141}, {"text": "speech recognition", "start": 156, "end": 174}, {"text": "packages", "start": 175, "end": 183}, {"text": "rules", "start": 217, "end": 222}, {"text": "transfer", "start": 240, "end": 248}, {"text": "semantics", "start": 267, "end": 276}, {"text": "unification", "start": 287, "end": 298}, {"text": "speech recognition", "start": 313, "end": 331}, {"text": "resulting", "start": 346, "end": 355}, {"text": "context-free", "start": 375, "end": 387}, {"text": "backbone", "start": 388, "end": 396}, {"text": "unification", "start": 404, "end": 415}, {"text": "grammar rules", "start": 485, "end": 498}, {"text": "method", "start": 504, "end": 510}, {"text": "tested", "start": 515, "end": 521}, {"text": "unification", "start": 540, "end": 551}, {"text": "English", "start": 564, "end": 571}, {"text": "speech recognition", "start": 585, "end": 603}, {"text": "software", "start": 604, "end": 612}, {"text": "corpus", "start": 618, "end": 624}, {"text": "utterances", "start": 632, "end": 642}, {"text": "Results", "start": 669, "end": 676}, {"text": "computational", "start": 699, "end": 712}, {"text": "respect", "start": 727, "end": 734}, {"text": "speech recognition", "start": 738, "end": 756}, {"text": "performances", "start": 757, "end": 769}, {"text": "speech recognition", "start": 774, "end": 792}, {"text": "semantics", "start": 820, "end": 829}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 104, "end": 109}, "tail": {"text": "method", "start": 112, "end": 118}}, {"head": {"text": "method", "start": 504, "end": 510}, "tail": {"text": "English", "start": 564, "end": 571}}, {"head": {"text": "software", "start": 604, "end": 612}, "tail": {"text": "corpus", "start": 618, "end": 624}}]}}, "schema": []} |
| {"input": "Feature Weighting For Co-Occurrence-Based Classification Of Words The paper comparatively studies methods of feature weighting in application to the task of cooccurrence-based classification of words according to their meaning. We explore parameter optimization of several weighting methods frequently used for similar problems such as text classification. We find that successful application of all the methods crucially depends on a number of parameters; only a carefully chosen weighting procedure allows to obtain consistent improvement on a classifier learned from non-weighteddata.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 70, "end": 75}, {"text": "studies", "start": 90, "end": 97}, {"text": "methods", "start": 98, "end": 105}, {"text": "feature", "start": 109, "end": 116}, {"text": "weighting", "start": 117, "end": 126}, {"text": "application", "start": 130, "end": 141}, {"text": "task", "start": 149, "end": 153}, {"text": "classification", "start": 176, "end": 190}, {"text": "words", "start": 194, "end": 199}, {"text": "parameter", "start": 239, "end": 248}, {"text": "optimization", "start": 249, "end": 261}, {"text": "weighting", "start": 273, "end": 282}, {"text": "methods", "start": 283, "end": 290}, {"text": "problems", "start": 319, "end": 327}, {"text": "text classification", "start": 336, "end": 355}, {"text": "application", "start": 381, "end": 392}, {"text": "methods", "start": 404, "end": 411}, {"text": "number", "start": 435, "end": 441}, {"text": "parameters", "start": 445, "end": 455}, {"text": "weighting", "start": 481, "end": 490}, {"text": "procedure", "start": 491, "end": 500}, {"text": "improvement", "start": 529, "end": 540}, {"text": "classifier", "start": 546, "end": 556}, {"text": "weighteddata", "start": 574, "end": 586}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 70, "end": 75}, "tail": {"text": "methods", "start": 98, "end": 105}}, {"head": {"text": "classification", "start": 176, "end": 190}, "tail": {"text": "words", "start": 194, "end": 199}}]}}, "schema": []} |
| {"input": "Trajectory Based Word Sense Disambiguation Classifier combination is a promising way to improve performance of word sense disambiguation. We propose a new combinational method in this paper. We first construct a series of Na ve Bayesian classifiers along a sequence of orderly varying sized windows of context, and perform sense selection for both training samples and test samples using these classifiers. We thus get a sense selection trajectory along the sequence of context windows for each sample. Then we make use of these trajectories to make final k-nearest-neighbors-based sense selection for test samples. This method aims to lower the uncertainty brought by classifiers using different context windows and make more robust utilization of context while perform well. Experiments show that our approach outperforms some other algorithms on both robustness and performance.", "output": {"entities": {"keyphrase": [{"text": "Classifier", "start": 43, "end": 53}, {"text": "combination", "start": 54, "end": 65}, {"text": "improve", "start": 88, "end": 95}, {"text": "performance", "start": 96, "end": 107}, {"text": "word sense disambiguation", "start": 111, "end": 136}, {"text": "propose", "start": 141, "end": 148}, {"text": "method", "start": 169, "end": 175}, {"text": "paper", "start": 184, "end": 189}, {"text": "construct", "start": 200, "end": 209}, {"text": "series", "start": 212, "end": 218}, {"text": "classifiers", "start": 237, "end": 248}, {"text": "sequence", "start": 257, "end": 265}, {"text": "sized", "start": 285, "end": 290}, {"text": "windows", "start": 291, "end": 298}, {"text": "context", "start": 302, "end": 309}, {"text": "perform", "start": 315, "end": 322}, {"text": "sense", "start": 323, "end": 328}, {"text": "selection", "start": 329, "end": 338}, {"text": "training", "start": 348, "end": 356}, {"text": "samples", "start": 357, "end": 364}, {"text": "test", "start": 369, "end": 373}, {"text": "samples", "start": 374, "end": 381}, {"text": "classifiers", "start": 394, "end": 405}, {"text": "sense", "start": 421, "end": 426}, {"text": "selection", "start": 427, "end": 436}, {"text": "sequence", "start": 458, "end": 466}, {"text": "context", "start": 470, "end": 477}, {"text": "windows", "start": 478, "end": 485}, {"text": "sample", "start": 495, "end": 501}, {"text": "sense", "start": 582, "end": 587}, {"text": "selection", "start": 588, "end": 597}, {"text": "test", "start": 602, "end": 606}, {"text": "samples", "start": 607, "end": 614}, {"text": "method", "start": 621, "end": 627}, {"text": "uncertainty", "start": 646, "end": 657}, {"text": "classifiers", "start": 669, "end": 680}, {"text": "context", "start": 697, "end": 704}, {"text": "windows", "start": 705, "end": 712}, {"text": "robust", "start": 727, "end": 733}, {"text": "context", "start": 749, "end": 756}, {"text": "perform", "start": 763, "end": 770}, {"text": "Experiments", "start": 777, "end": 788}, {"text": "approach", "start": 803, "end": 811}, {"text": "algorithms", "start": 835, "end": 845}, {"text": "robustness", "start": 854, "end": 864}, {"text": "performance", "start": 869, "end": 880}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 184, "end": 189}, "tail": {"text": "method", "start": 169, "end": 175}}, {"head": {"text": "classifiers", "start": 394, "end": 405}, "tail": {"text": "selection", "start": 329, "end": 338}}, {"head": {"text": "windows", "start": 705, "end": 712}, "tail": {"text": "method", "start": 621, "end": 627}}]}}, "schema": []} |
| {"input": "Automated Induction Of Sense In Context In this paper, we introduce a model for sense assignment which relies on assigning senses to the contexts within which words appear, rather than to the words themselves. We argue that word senses as such are not directly encoded in the lexicon of the language. Rather, each word is associated with one or more stereotypical syntagmatic patterns, which we call selection contexts.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 48, "end": 53}, {"text": "model", "start": 70, "end": 75}, {"text": "sense", "start": 80, "end": 85}, {"text": "assignment", "start": 86, "end": 96}, {"text": "contexts", "start": 137, "end": 145}, {"text": "words", "start": 159, "end": 164}, {"text": "words", "start": 192, "end": 197}, {"text": "word", "start": 224, "end": 228}, {"text": "lexicon", "start": 276, "end": 283}, {"text": "language", "start": 291, "end": 299}, {"text": "word", "start": 314, "end": 318}, {"text": "patterns", "start": 376, "end": 384}, {"text": "call", "start": 395, "end": 399}, {"text": "selection", "start": 400, "end": 409}, {"text": "contexts", "start": 410, "end": 418}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 48, "end": 53}, "tail": {"text": "model", "start": 70, "end": 75}}, {"head": {"text": "contexts", "start": 137, "end": 145}, "tail": {"text": "assignment", "start": 86, "end": 96}}]}}, "schema": []} |
| {"input": "Situations And Prepositional Phrases This paper presents a format for representing the linguistic form of utterances, called situation schemata, which is rooted in the situation semantics of Barwise and Perry. A treatment of locative prepositional phrases is given, thus illustrating the generation of the situation schemata and their interpretation in situation semantics.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 42, "end": 47}, {"text": "format", "start": 59, "end": 65}, {"text": "form", "start": 98, "end": 102}, {"text": "utterances", "start": 106, "end": 116}, {"text": "called", "start": 118, "end": 124}, {"text": "situation", "start": 125, "end": 134}, {"text": "schemata", "start": 135, "end": 143}, {"text": "situation", "start": 168, "end": 177}, {"text": "semantics", "start": 178, "end": 187}, {"text": "treatment", "start": 212, "end": 221}, {"text": "prepositional phrases", "start": 234, "end": 255}, {"text": "generation", "start": 288, "end": 298}, {"text": "situation", "start": 306, "end": 315}, {"text": "schemata", "start": 316, "end": 324}, {"text": "interpretation", "start": 335, "end": 349}, {"text": "situation", "start": 353, "end": 362}, {"text": "semantics", "start": 363, "end": 372}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 42, "end": 47}, "tail": {"text": "format", "start": 59, "end": 65}}, {"head": {"text": "form", "start": 98, "end": 102}, "tail": {"text": "utterances", "start": 106, "end": 116}}, {"head": {"text": "interpretation", "start": 335, "end": 349}, "tail": {"text": "schemata", "start": 316, "end": 324}}]}}, "schema": []} |
| {"input": "The U. S. Policy Agenda Legislation Corpus Volume 1-a Language Resource from 1947-1998 We introduce the corpus of United States Congressional bills from 1947 to 1998 for use by language research communities. The U. S. Policy Agenda Legislation Corpus Volume 1 (USPALCV1) includes more than 375, 000 legislative bills annotated with a hierarchical policy area category. The human annotations in USPALCV1 have been reliably applied over time to enable social science analysis of legislative trends. The corpus is a member of an emerging family of corpora that are annotated by policy area to enable comparative parallel trend recognition across countries and domains (legislation, political speeches, newswire articles, budgetary expenditures, web sites, etc.). This paper describes the origins of the corpus, its creation, ways to access it, design criteria, and an analysis with common supervised machine learning methods. The use of machine learning methods establishes a baseline proposed modeling for the topic classification of legal documents.", "output": {"entities": {"keyphrase": [{"text": "corpus", "start": 104, "end": 110}, {"text": "language", "start": 177, "end": 185}, {"text": "research", "start": 186, "end": 194}, {"text": "communities", "start": 195, "end": 206}, {"text": "Agenda", "start": 225, "end": 231}, {"text": "Corpus", "start": 244, "end": 250}, {"text": "Volume", "start": 251, "end": 257}, {"text": "includes", "start": 271, "end": 279}, {"text": "area", "start": 354, "end": 358}, {"text": "category", "start": 359, "end": 367}, {"text": "applied", "start": 422, "end": 429}, {"text": "time", "start": 435, "end": 439}, {"text": "science", "start": 457, "end": 464}, {"text": "analysis", "start": 465, "end": 473}, {"text": "trends", "start": 489, "end": 495}, {"text": "corpus", "start": 501, "end": 507}, {"text": "corpora", "start": 545, "end": 552}, {"text": "area", "start": 582, "end": 586}, {"text": "trend", "start": 618, "end": 623}, {"text": "recognition", "start": 624, "end": 635}, {"text": "domains", "start": 657, "end": 664}, {"text": "speeches", "start": 689, "end": 697}, {"text": "web sites", "start": 742, "end": 751}, {"text": "paper", "start": 765, "end": 770}, {"text": "corpus", "start": 800, "end": 806}, {"text": "creation", "start": 812, "end": 820}, {"text": "access", "start": 830, "end": 836}, {"text": "design", "start": 841, "end": 847}, {"text": "criteria", "start": 848, "end": 856}, {"text": "analysis", "start": 865, "end": 873}, {"text": "common", "start": 879, "end": 885}, {"text": "machine", "start": 897, "end": 904}, {"text": "methods", "start": 914, "end": 921}, {"text": "machine", "start": 934, "end": 941}, {"text": "methods", "start": 951, "end": 958}, {"text": "proposed", "start": 982, "end": 990}, {"text": "modeling", "start": 991, "end": 999}, {"text": "topic", "start": 1008, "end": 1013}, {"text": "classification", "start": 1014, "end": 1028}, {"text": "documents", "start": 1038, "end": 1047}]}, "relations": {"no_relation": [{"head": {"text": "corpus", "start": 104, "end": 110}, "tail": {"text": "research", "start": 186, "end": 194}}, {"head": {"text": "corpus", "start": 501, "end": 507}, "tail": {"text": "corpora", "start": 545, "end": 552}}, {"head": {"text": "paper", "start": 765, "end": 770}, "tail": {"text": "corpus", "start": 800, "end": 806}}, {"head": {"text": "classification", "start": 1014, "end": 1028}, "tail": {"text": "documents", "start": 1038, "end": 1047}}]}}, "schema": []} |
| {"input": "JMWNL: an Extensible Multilingual Library for Accessing Wordnets in Different Languages In this paper we present JMWNL, a multilingual extension of the JWNL java library, which was originally developed for accessing Princeton WordNet dictionaries. JMWNL broadens the range of JWNL' s accessible resources by covering also dictionaries produced inside the EuroWordNet project. Specific resources, such as language-dependent algorithmic stemmers, have been adopted to cover the diversities in the morphological nature of words in the addressed idioms. New semantic and lexical relations have been included to maximize compatibility with new versions of the original Princeton WordNet and to include the whole range of relations from EuroWordNet. Relations from Princeton WordNet on one side and EuroWordNet on the other one have in some cases been mapped to provide a uniform reference for coherent cross-linguistic use of the library.outputentitieskeyphrasetextpaperstartendtextextensionstartendtextlibrarystartendtextdevelopedstartendtextaccessingstartendtextdictionariesstartendtextresourcesstartendtextdictionariesstartendtextprojectstartendtextresourcesstartendtextlanguage-dependentstartendtextdiversitiesstartendtextnaturestartendtextwordsstartendtextidiomsstartendtextsemanticstartendtextlexicalstartendtextrelationsstartendtextincludedstartendtextcompatibilitystartendtextversionsstartendtextincludestartendtextrelationsstartendtextRelationsstartendtextsidestartendtextcasesstartendtextmappedstartendtextprovidestartendtextreferencestartendtextcross-linguisticstartendtextlibrarystartendrelationsno_relationheadtextpaperstartendtailtextextensionstartendheadtextdiversitiesstartendtailtextwordsstartendschema |
| inputMethodology for Evaluating the Usability of User Interfaces in Mobile Services In this paper we present a usability measure adapted to mobile services, which is based on the well-known theoretical framework defined in the ISO 9241-11 [ISO 9241 (1988)] standard. This measure is then applied to a representative set of services of the Telefonica' s portfolio for residential customers. The user tests that we present were carried out by a total of 327 people. Additionally, in section 3 we describe the application of the methodology to a particular service and section 4 presents the results of the experiments. These results show highly significant differences in the three usability measures considered, though all of them have the same trend. The worst performers in all cases were the WAP and i-mode user interfaces (UI), while the best performers were the SMS and web based UIs closely followed by the voice UI. Finally, in section 5 we analyse the results and present our conclusions.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 87, "end": 92}, {"text": "adapted", "start": 124, "end": 131}, {"text": "based", "start": 161, "end": 166}, {"text": "framework", "start": 197, "end": 206}, {"text": "standard", "start": 252, "end": 260}, {"text": "applied", "start": 283, "end": 290}, {"text": "representative", "start": 296, "end": 310}, {"text": "customers", "start": 374, "end": 383}, {"text": "user", "start": 389, "end": 393}, {"text": "tests", "start": 394, "end": 399}, {"text": "section", "start": 476, "end": 483}, {"text": "application", "start": 502, "end": 513}, {"text": "methodology", "start": 521, "end": 532}, {"text": "section", "start": 561, "end": 568}, {"text": "results", "start": 584, "end": 591}, {"text": "experiments", "start": 599, "end": 610}, {"text": "results", "start": 618, "end": 625}, {"text": "differences", "start": 650, "end": 661}, {"text": "trend", "start": 739, "end": 744}, {"text": "cases", "start": 774, "end": 779}, {"text": "i-mode", "start": 797, "end": 803}, {"text": "user interfaces", "start": 804, "end": 819}, {"text": "based", "start": 873, "end": 878}, {"text": "section", "start": 929, "end": 936}, {"text": "results", "start": 954, "end": 961}, {"text": "conclusions", "start": 978, "end": 989}]}, "relations": {"no_relation": [{"head": {"text": "experiments", "start": 599, "end": 610}, "tail": {"text": "results", "start": 584, "end": 591}}]}}, "schema": []} |
| {"input": "The 2008 Oriental COCOSDA Book Project: in Commemoration of the First Decade of Sustained Activities in Asia The purpose of Oriental COCOSDA is to provide the Asian community a platform to exchange ideas, to share information and to discuss regional matters on creation, utilization, dissemination of spoken language corpora of oriental languages and also on the assessment methods of speech recognition/synthesis systems as well as to promote speech research on oriental languages. Since its preparatory meeting in Hong Kong in 1997, annual workshops have been organized and held in Japan, Taiwan, China, Korea, Thailand, Singapore, India, Indonesia, Malaysia, and Vietnam from 1998. The organization is managed by a convener, three advisory members, and 26 committee members from 13 regions in Oriental area. In order to commemorate 10 years of continued activities, the members have decided to publish a book which covers a wide range of speech research. Special focus will be on speech resources or speech corpora in Oriental countries and standardization of speech input/output systems performance evaluation methods on which key technologies for speech systems development are based. The book will also include linguistic outlines of oriental languages, annotation, labeling, and software tools for speech processing.", "output": {"entities": {"keyphrase": [{"text": "purpose", "start": 113, "end": 120}, {"text": "provide", "start": 147, "end": 154}, {"text": "community", "start": 165, "end": 174}, {"text": "platform", "start": 177, "end": 185}, {"text": "information", "start": 214, "end": 225}, {"text": "creation", "start": 261, "end": 269}, {"text": "language", "start": 308, "end": 316}, {"text": "corpora", "start": 317, "end": 324}, {"text": "languages", "start": 337, "end": 346}, {"text": "assessment", "start": 363, "end": 373}, {"text": "methods", "start": 374, "end": 381}, {"text": "speech recognition", "start": 385, "end": 403}, {"text": "synthesis", "start": 404, "end": 413}, {"text": "systems", "start": 414, "end": 421}, {"text": "speech", "start": 444, "end": 450}, {"text": "research", "start": 451, "end": 459}, {"text": "languages", "start": 472, "end": 481}, {"text": "meeting", "start": 505, "end": 512}, {"text": "Hong", "start": 516, "end": 520}, {"text": "workshops", "start": 542, "end": 551}, {"text": "organization", "start": 689, "end": 701}, {"text": "regions", "start": 785, "end": 792}, {"text": "area", "start": 805, "end": 809}, {"text": "order", "start": 814, "end": 819}, {"text": "speech", "start": 941, "end": 947}, {"text": "research", "start": 948, "end": 956}, {"text": "focus", "start": 966, "end": 971}, {"text": "speech", "start": 983, "end": 989}, {"text": "resources", "start": 990, "end": 999}, {"text": "speech corpora", "start": 1003, "end": 1017}, {"text": "speech", "start": 1063, "end": 1069}, {"text": "input", "start": 1070, "end": 1075}, {"text": "output", "start": 1076, "end": 1082}, {"text": "systems", "start": 1083, "end": 1090}, {"text": "performance", "start": 1091, "end": 1102}, {"text": "evaluation methods", "start": 1103, "end": 1121}, {"text": "technologies", "start": 1135, "end": 1147}, {"text": "speech", "start": 1152, "end": 1158}, {"text": "systems", "start": 1159, "end": 1166}, {"text": "development", "start": 1167, "end": 1178}, {"text": "based", "start": 1183, "end": 1188}, {"text": "include", "start": 1209, "end": 1216}, {"text": "outlines", "start": 1228, "end": 1236}, {"text": "languages", "start": 1249, "end": 1258}, {"text": "software", "start": 1286, "end": 1294}, {"text": "tools", "start": 1295, "end": 1300}, {"text": "speech processing", "start": 1305, "end": 1322}]}, "relations": {"no_relation": [{"head": {"text": "languages", "start": 337, "end": 346}, "tail": {"text": "corpora", "start": 317, "end": 324}}, {"head": {"text": "methods", "start": 374, "end": 381}, "tail": {"text": "assessment", "start": 363, "end": 373}}, {"head": {"text": "tools", "start": 1295, "end": 1300}, "tail": {"text": "speech processing", "start": 1305, "end": 1322}}]}}, "schema": []} |
| {"input": "A CD-ROM Retrieval System With Multiple Dialogue Agents In this paper, we proposed a new dialogue system with multiple dialogue agents. In our new system, three types of agents: a) domain agents, b) strategy agents, and c) context agents were realized. They give the following advantages to the user: the domain agents make the user aware of the boundary between the domains. the strategy agents make the user aware of the difference between the strategies. the context agents help the user to deal with multiple goals. We expect that the complex behaviors of the system will become more visible to the user in different situations. The experimental results show that the user can retrieve effectively and obtain the expected goals easily by using these multiple agents.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 64, "end": 69}, {"text": "proposed", "start": 74, "end": 82}, {"text": "dialogue system", "start": 89, "end": 104}, {"text": "dialogue", "start": 119, "end": 127}, {"text": "agents", "start": 128, "end": 134}, {"text": "system", "start": 147, "end": 153}, {"text": "types", "start": 161, "end": 166}, {"text": "agents", "start": 170, "end": 176}, {"text": "domain", "start": 181, "end": 187}, {"text": "agents", "start": 188, "end": 194}, {"text": "strategy", "start": 199, "end": 207}, {"text": "agents", "start": 208, "end": 214}, {"text": "context", "start": 223, "end": 230}, {"text": "agents", "start": 231, "end": 237}, {"text": "advantages", "start": 277, "end": 287}, {"text": "user", "start": 295, "end": 299}, {"text": "domain", "start": 305, "end": 311}, {"text": "agents", "start": 312, "end": 318}, {"text": "user", "start": 328, "end": 332}, {"text": "boundary", "start": 346, "end": 354}, {"text": "domains", "start": 367, "end": 374}, {"text": "strategy", "start": 380, "end": 388}, {"text": "agents", "start": 389, "end": 395}, {"text": "user", "start": 405, "end": 409}, {"text": "difference", "start": 423, "end": 433}, {"text": "strategies", "start": 446, "end": 456}, {"text": "context", "start": 462, "end": 469}, {"text": "agents", "start": 470, "end": 476}, {"text": "help", "start": 477, "end": 481}, {"text": "user", "start": 486, "end": 490}, {"text": "deal", "start": 494, "end": 498}, {"text": "goals", "start": 513, "end": 518}, {"text": "complex", "start": 539, "end": 546}, {"text": "behaviors", "start": 547, "end": 556}, {"text": "system", "start": 564, "end": 570}, {"text": "user", "start": 603, "end": 607}, {"text": "situations", "start": 621, "end": 631}, {"text": "experimental", "start": 637, "end": 649}, {"text": "results", "start": 650, "end": 657}, {"text": "user", "start": 672, "end": 676}, {"text": "goals", "start": 726, "end": 731}, {"text": "agents", "start": 763, "end": 769}]}, "relations": {"no_relation": [{"head": {"text": "paper", "start": 64, "end": 69}, "tail": {"text": "dialogue system", "start": 89, "end": 104}}]}}, "schema": []} |
| {"input": "Computation Of Relative Social Status On The Basis Of Honorification In Korean This paper presents a way to compute relative social status of the individuals involved in Korean dialogue. Every Korean sentence indicates whether honorification occurs in it. The occurrence of honorification in a sentence is constr ained by relative social status of the individuals involved in the sentence. By using the information about social status and the information about sentence-external individuals such as speaker and addressee, we can explain why a sentence is felicitous in a restricted context and whether a dialogue is coherent or not. Since it is possible and easy to include such contextual information in the HPSG formalism, that formalism is adopted here. The implementation of Korean dialogue processing and the computation of social status is made based on ALE system.", "output": {"entities": {"keyphrase": [{"text": "paper", "start": 84, "end": 89}, {"text": "compute", "start": 108, "end": 115}, {"text": "relative", "start": 116, "end": 124}, {"text": "status", "start": 132, "end": 138}, {"text": "individuals", "start": 146, "end": 157}, {"text": "dialogue", "start": 177, "end": 185}, {"text": "sentence", "start": 200, "end": 208}, {"text": "occurrence", "start": 260, "end": 270}, {"text": "sentence", "start": 294, "end": 302}, {"text": "relative", "start": 322, "end": 330}, {"text": "status", "start": 338, "end": 344}, {"text": "individuals", "start": 352, "end": 363}, {"text": "sentence", "start": 380, "end": 388}, {"text": "information", "start": 403, "end": 414}, {"text": "status", "start": 428, "end": 434}, {"text": "information", "start": 443, "end": 454}, {"text": "sentence-external", "start": 461, "end": 478}, {"text": "individuals", "start": 479, "end": 490}, {"text": "sentence", "start": 543, "end": 551}, {"text": "context", "start": 582, "end": 589}, {"text": "dialogue", "start": 604, "end": 612}, {"text": "include", "start": 666, "end": 673}, {"text": "contextual information", "start": 679, "end": 701}, {"text": "formalism", "start": 714, "end": 723}, {"text": "formalism", "start": 730, "end": 739}, {"text": "implementation", "start": 761, "end": 775}, {"text": "dialogue", "start": 786, "end": 794}, {"text": "processing", "start": 795, "end": 805}, {"text": "computation", "start": 814, "end": 825}, {"text": "status", "start": 836, "end": 842}, {"text": "based", "start": 851, "end": 856}, {"text": "system", "start": 864, "end": 870}]}, "relations": {"no_relation": [{"head": {"text": "contextual information", "start": 679, "end": 701}, "tail": {"text": "formalism", "start": 714, "end": 723}}, {"head": {"text": "processing", "start": 795, "end": 805}, "tail": {"text": "dialogue", "start": 786, "end": 794}}]}}, "schema": []} |
| {"input": "Learning Part-Of-Speech Guessing Rules From Lexicon: Extension To Non-Concatenative Operations One of the problems in part-of-speech tagging of real-word texts is that of unknown to the lexicon words. In (Mikheev, 1996), a technique for fully unsupervised statistical acquisition of rules which guess possible parts-of-speech for unknown words was proposed. One of the over-simplification assumed by this learning technique was the acquisition of morphological rules which obey only simple concatenative regularities of the main word with an affix. In this paper wc extend this technique to the non-concatenative cases of suffixation and assess the gain in the performance.", "output": {"entities": {"keyphrase": [{"text": "problems", "start": 106, "end": 114}, {"text": "part-of-speech", "start": 118, "end": 132}, {"text": "tagging", "start": 133, "end": 140}, {"text": "real-word", "start": 144, "end": 153}, {"text": "texts", "start": 154, "end": 159}, {"text": "lexicon", "start": 186, "end": 193}, {"text": "words", "start": 194, "end": 199}, {"text": "technique", "start": 223, "end": 232}, {"text": "statistical", "start": 256, "end": 267}, {"text": "acquisition", "start": 268, "end": 279}, {"text": "rules", "start": 283, "end": 288}, {"text": "parts-of-speech", "start": 310, "end": 325}, {"text": "unknown words", "start": 330, "end": 343}, {"text": "proposed", "start": 348, "end": 356}, {"text": "over-simplification", "start": 369, "end": 388}, {"text": "technique", "start": 414, "end": 423}, {"text": "acquisition", "start": 432, "end": 443}, {"text": "rules", "start": 461, "end": 466}, {"text": "simple", "start": 483, "end": 489}, {"text": "regularities", "start": 504, "end": 516}, {"text": "main", "start": 524, "end": 528}, {"text": "word", "start": 529, "end": 533}, {"text": "paper", "start": 557, "end": 562}, {"text": "technique", "start": 578, "end": 587}, {"text": "cases", "start": 613, "end": 618}, {"text": "gain", "start": 649, "end": 653}, {"text": "performance", "start": 661, "end": 672}]}, "relations": {"no_relation": [{"head": {"text": "tagging", "start": 133, "end": 140}, "tail": {"text": "texts", "start": 154, "end": 159}}, {"head": {"text": "technique", "start": 223, "end": 232}, "tail": {"text": "acquisition", "start": 268, "end": 279}}, {"head": {"text": "parts-of-speech", "start": 310, "end": 325}, "tail": {"text": "unknown words", "start": 330, "end": 343}}, {"head": {"text": "paper", "start": 557, "end": 562}, "tail": {"text": "technique", "start": 578, "end": 587}}]}}, "schema": []} |
| {"input": "An Efficient Execution Method For Rule-Based Machine Translation A rule based system is an effective way to implement a machine translation system because of its extensibility and maintainability. However, it is disadvantageous in processing efficiency. In a rule based machine translation system, the grammar consists of a lot of rewriting rules. While the translation is carried out by repeating pattern matching and transformation of graph structures, most rules fail in pattern matching. It is to be desired that pattern matching of the unfruitful rules should be avoided. This paper proposes a method to restrict the rule application by activating rules dynamically. The logical relationship among rules are pre-analyzed and a set of antecedent actions, which are prerequisite for the condition of the rule being satisfied, is determined for each rule. In execution time, a rule is activated only when one of the antecedent actions are carried out. The probability of a rule being activated is reduced to near the occurrence probability of its relevant linguistic phenomenon. As most rules relate to linguistic phenomena that rarely occur, the processing efficiency is drastically improved.", "output": {"entities": {"keyphrase": [{"text": "rule", "start": 67, "end": 71}, {"text": "based", "start": 72, "end": 77}, {"text": "system", "start": 78, "end": 84}, {"text": "implement", "start": 108, "end": 117}, {"text": "machine translation system", "start": 120, "end": 146}, {"text": "processing", "start": 231, "end": 241}, {"text": "efficiency", "start": 242, "end": 252}, {"text": "rule", "start": 259, "end": 263}, {"text": "based", "start": 264, "end": 269}, {"text": "machine translation system", "start": 270, "end": 296}, {"text": "rules", "start": 341, "end": 346}, {"text": "translation", "start": 358, "end": 369}, {"text": "pattern matching", "start": 398, "end": 414}, {"text": "transformation", "start": 419, "end": 433}, {"text": "structures", "start": 443, "end": 453}, {"text": "rules", "start": 460, "end": 465}, {"text": "pattern matching", "start": 474, "end": 490}, {"text": "pattern matching", "start": 517, "end": 533}, {"text": "rules", "start": 552, "end": 557}, {"text": "paper", "start": 582, "end": 587}, {"text": "proposes", "start": 588, "end": 596}, {"text": "method", "start": 599, "end": 605}, {"text": "rule application", "start": 622, "end": 638}, {"text": "rules", "start": 653, "end": 658}, {"text": "relationship", "start": 684, "end": 696}, {"text": "rules", "start": 703, "end": 708}, {"text": "actions", "start": 750, "end": 757}, {"text": "rule", "start": 807, "end": 811}, {"text": "rule", "start": 852, "end": 856}, {"text": "execution", "start": 861, "end": 870}, {"text": "time", "start": 871, "end": 875}, {"text": "rule", "start": 879, "end": 883}, {"text": "actions", "start": 929, "end": 936}, {"text": "probability", "start": 958, "end": 969}, {"text": "rule", "start": 975, "end": 979}, {"text": "occurrence", "start": 1019, "end": 1029}, {"text": "probability", "start": 1030, "end": 1041}, {"text": "phenomenon", "start": 1069, "end": 1079}, {"text": "rules", "start": 1089, "end": 1094}, {"text": "phenomena", "start": 1116, "end": 1125}, {"text": "processing", "start": 1149, "end": 1159}, {"text": "efficiency", "start": 1160, "end": 1170}, {"text": "improved", "start": 1186, "end": 1194}]}, "relations": {"no_relation": [{"head": {"text": "system", "start": 78, "end": 84}, "tail": {"text": "machine translation system", "start": 120, "end": 146}}, {"head": {"text": "rules", "start": 341, "end": 346}, "tail": {"text": "machine translation system", "start": 270, "end": 296}}, {"head": {"text": "paper", "start": 582, "end": 587}, "tail": {"text": "method", "start": 599, "end": 605}}, {"head": {"text": "probability", "start": 958, "end": 969}, "tail": {"text": "rule", "start": 975, "end": 979}}, {"head": {"text": "probability", "start": 1030, "end": 1041}, "tail": {"text": "phenomenon", "start": 1069, "end": 1079}}, {"head": {"text": "rules", "start": 1089, "end": 1094}, "tail": {"text": "phenomena", "start": 1116, "end": 1125}}]}}, "schema": []} |
| {"input": "Hopfield Models As Nondeterministic Finite-State Machines The use of neural networks for integrated linguistic analysis may be profitable. This paper presents the first results of our research on that subject: a Hop-field model for syntactical analysis. We construct a neural network as an implementation of a bounded push-down automaton, which can accept context-free languages with limited center-embedding. The network' s behavior can be predicted a priori, so the presented theory can be tested. The operation of the network as an implementation of the acceptor is prov-ably correct. Furthermore we found a solution to the problem of spurious states in Hopfield models: we use them as dynamically constructed representations of sets of states of the implemented acceptor. The so-called neural-network acceptor we propose, is fast but large.outputentitieskeyphrasetextModelsstartendtextneural networksstartendtextlinguistic analysisstartendtextpaperstartendtextresultsstartendtextresearchstartendtextHop-fieldstartendtextmodelstartendtextanalysisstartendtextconstructstartendtextneural networkstartendtextimplementationstartendtextcontext-freestartendtextlanguagesstartendtextcenter-embeddingstartendtextnetworkstartendtextbehaviorstartendtexttheorystartendtexttestedstartendtextoperationstartendtextnetworkstartendtextimplementationstartendtextsolutionstartendtextproblemstartendtextmodelsstartendtextconstructedstartendtextrepresentationsstartendtextimplementedstartendtextneural-networkstartendtextproposestartendrelationsno_relationheadtextneural networksstartendtailtextlinguistic analysisstartendheadtextmodelstartendtailtextanalysisstartendschema |
| inputGeneration Of Accent In Nominally Premodified Noun Phrases The primary purpose of this paper is to present a set of conditions that constrain accent placement in focused nominally premodified Selkirk (1984) argues that if the premodifier is an argument of the head, then the head can be deaccented. I agree with Selkirk' s proposal and argue that what essential is not whether the premodifier is a grammatical argument of the head noun, but rather, whether it is a 6-complement in lexical conceptual structure. This proposal is evaluated by testing it against a corpus of naturally occurringdata.", "output": {"entities": {"keyphrase": [{"text": "purpose", "start": 71, "end": 78}, {"text": "paper", "start": 87, "end": 92}, {"text": "focused", "start": 162, "end": 169}, {"text": "argument", "start": 244, "end": 252}, {"text": "proposal", "start": 323, "end": 331}, {"text": "argument", "start": 410, "end": 418}, {"text": "noun", "start": 431, "end": 435}, {"text": "complement", "start": 467, "end": 477}, {"text": "lexical", "start": 481, "end": 488}, {"text": "conceptual structure", "start": 489, "end": 509}, {"text": "proposal", "start": 516, "end": 524}, {"text": "evaluated", "start": 528, "end": 537}, {"text": "testing", "start": 541, "end": 548}, {"text": "corpus", "start": 562, "end": 568}, {"text": "occurringdata", "start": 582, "end": 595}]}, "relations": {"no_relation": [{"head": {"text": "occurringdata", "start": 582, "end": 595}, "tail": {"text": "corpus", "start": 562, "end": 568}}]}}, "schema": []} |
| |
