File size: 85,772 Bytes
6fa4bc9 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 | {
"paper_id": "C08-1012",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T12:25:23.428361Z"
},
"title": "Are Morpho-Syntactic Features More Predictive for the Resolution of Noun Phrase Coordination Ambiguity than Lexico-Semantic Similarity Scores?",
"authors": [
{
"first": "Ekaterina",
"middle": [],
"last": "Buyko",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jena University Language & Information Engineering (JULIE)",
"location": {
"addrLine": "Lab F\u00fcrstengraben 30",
"postCode": "07743",
"settlement": "Jena",
"country": "Germany"
}
},
"email": "ekaterina.buyko|udo.hahn@uni-jena.de"
},
{
"first": "Udo",
"middle": [],
"last": "Hahn",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jena University Language & Information Engineering (JULIE)",
"location": {
"addrLine": "Lab F\u00fcrstengraben 30",
"postCode": "07743",
"settlement": "Jena",
"country": "Germany"
}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Coordinations in noun phrases often pose the problem that elliptified parts have to be reconstructed for proper semantic interpretation. Unfortunately, the detection of coordinated heads and identification of elliptified elements notoriously lead to ambiguous reconstruction alternatives. While linguistic intuition suggests that semantic criteria might play an important, if not superior, role in disambiguating resolution alternatives, our experiments on the reannotated WSJ part of the Penn Treebank indicate that solely morpho-syntactic criteria are more predictive than solely lexicosemantic ones. We also found that the combination of both criteria does not yield any substantial improvement. 1 Introduction Looking at noun phrases such as 'cat and dog owner' 'novels and travel books' their proper coordination reading (and asymmetric distribution of coordinated heads) as 'cat owner' AND 'dog owner' 'novels' AND 'travel books' seems to be licensed by the striking semantic similarity between 'cat' and 'dog', and 'novels' and 'books', respectively. If this were a general rule, then automatic procedures for the resolution of coordination ambiguities had to rely on the a priori provision of potentially large amounts of semantic",
"pdf_parse": {
"paper_id": "C08-1012",
"_pdf_hash": "",
"abstract": [
{
"text": "Coordinations in noun phrases often pose the problem that elliptified parts have to be reconstructed for proper semantic interpretation. Unfortunately, the detection of coordinated heads and identification of elliptified elements notoriously lead to ambiguous reconstruction alternatives. While linguistic intuition suggests that semantic criteria might play an important, if not superior, role in disambiguating resolution alternatives, our experiments on the reannotated WSJ part of the Penn Treebank indicate that solely morpho-syntactic criteria are more predictive than solely lexicosemantic ones. We also found that the combination of both criteria does not yield any substantial improvement. 1 Introduction Looking at noun phrases such as 'cat and dog owner' 'novels and travel books' their proper coordination reading (and asymmetric distribution of coordinated heads) as 'cat owner' AND 'dog owner' 'novels' AND 'travel books' seems to be licensed by the striking semantic similarity between 'cat' and 'dog', and 'novels' and 'books', respectively. If this were a general rule, then automatic procedures for the resolution of coordination ambiguities had to rely on the a priori provision of potentially large amounts of semantic",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "background knowledge to make this similarity explicit. Furthermore, any changes in languages or domains where such resources were missing (or, were incomplete) would severely hamper coordination analysis.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Indeed, previous research has gathered lot of evidence that conjoined elements tend to be semantically similar. The important role of semantic similarity criteria for properly sorting out conjuncts was first tested by Resnik (1999) . He introduced an information-content-based similarity measure that uses WORDNET (Fellbaum, 1998) as a lexico-semantic resource and came up with the claim that semantic similarity is helpful to achieve higher coverage in coordination resolution for coordinated noun phrases of the form 'noun1 and noun2 noun3' than similarity measures based on morphological information only.",
"cite_spans": [
{
"start": 218,
"end": 231,
"text": "Resnik (1999)",
"ref_id": "BIBREF20"
},
{
"start": 314,
"end": 330,
"text": "(Fellbaum, 1998)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "In a similar vein, Hogan (2007b) inspected WORDNET similarity and relatedness measures and investigated their role in conjunct identification. Her data reveals that several measures of semantic word similarity can indeed detect conjunct similarity. For the majority of these similarity measures, the differences between the mean similarity of coordinated elements and noncoordinated ones were statistically significant. However, it also became evident that these were only slight differences, and not all coordinated heads were semantically related as evidenced, e.g., by 'work '/'harmony' in 'hard work and harmony'. The significance tests did also not reveal particularly useful measures for conjunct identification. Rus et al. (2002) in an earlier study presented an alternative heuristics-based approach to conjunct identification for coordinations of the form 'noun1 and noun2 noun3'. They exploit, e.g., look-ups in WORDNET for a compound noun as a con-cept, and for the sibling relation between nouns in the coordination and report bracketing precision of 87.4% on 525 candidate coordinations. Although the authors demonstrated that WORDNET was really helpful in coordination resolution, the evaluation was only conducted on compound nouns extracted from WORDNET's noun hierachy and, furthermore, the senses of nouns were manually tagged in advance for the experiments.",
"cite_spans": [
{
"start": 19,
"end": 32,
"text": "Hogan (2007b)",
"ref_id": "BIBREF9"
},
{
"start": 578,
"end": 617,
"text": "'/'harmony' in 'hard work and harmony'.",
"ref_id": null
},
{
"start": 719,
"end": 736,
"text": "Rus et al. (2002)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Despite this preference for semantic criteria, one might still raise the question how far non-semantic criteria might guide the resolution of noun phrase coordination ambiguities, e.g., by means of the distribution of resolution alternatives in a large corpus or plain lexical or morpho-syntactic criteria. This idea has already been explored before by various researchers from different methodological angles including distribution-based statistical approaches (e.g., Chantree et al. (2005) , Nakov and Hearst (2005) ), similarity-based approaches incorporating orthographical, morpho-syntactic, and syntactic similarity criteria (e.g., Agarwal and Boggess (1992) , Okumura and Muraki (1994) ), as well as a combination of distribution information and syntactic criteria (Hogan, 2007a) .",
"cite_spans": [
{
"start": 469,
"end": 491,
"text": "Chantree et al. (2005)",
"ref_id": "BIBREF4"
},
{
"start": 494,
"end": 517,
"text": "Nakov and Hearst (2005)",
"ref_id": "BIBREF16"
},
{
"start": 638,
"end": 664,
"text": "Agarwal and Boggess (1992)",
"ref_id": "BIBREF0"
},
{
"start": 667,
"end": 692,
"text": "Okumura and Muraki (1994)",
"ref_id": "BIBREF18"
},
{
"start": 772,
"end": 786,
"text": "(Hogan, 2007a)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Statistical approaches enumerate all candidate conjuncts and calculate the respective likelihood according to a distribution estimated on a corpus. For the coordination 'movie and television industry' the distributional similarity of 'movie' and 'industry' and the collocation frequencies of the pairs ['movie' -'industry'] and ['television' -'industry'] would be compared against each other. However, for such an approach only an F-measure under 50% was reported (Chantree et al., 2005) . Unsupervised Web-distribution-based algorithms (Nakov and Hearst, 2005) achieved 80% on the disambiguation of coordinations of the fixed form 'noun1 and noun2 noun3'. Hogan (2007a) presented a method for the disambiguation of noun phrase coordination by modelling two sources of information, viz. distribution-based similarity between conjuncts and the dependency between conjunct heads. This method was incorporated in Bikel's parsing model (Bikel, 2004) and achieved an increase in NP coordination dependency Fscore from 69.9% to 73.8%.",
"cite_spans": [
{
"start": 302,
"end": 323,
"text": "['movie' -'industry']",
"ref_id": null
},
{
"start": 328,
"end": 354,
"text": "['television' -'industry']",
"ref_id": null
},
{
"start": 464,
"end": 487,
"text": "(Chantree et al., 2005)",
"ref_id": "BIBREF4"
},
{
"start": 537,
"end": 561,
"text": "(Nakov and Hearst, 2005)",
"ref_id": "BIBREF16"
},
{
"start": 657,
"end": 670,
"text": "Hogan (2007a)",
"ref_id": "BIBREF8"
},
{
"start": 932,
"end": 945,
"text": "(Bikel, 2004)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Similarity-based approaches consider those elements of a coordination as conjuncts which are most 'similar' under syntactic, morphological, or even semantic aspects. Agarwal and Boggess (1992) include in their NP coordination analysis syntactic and some semantic information about candidate conjuncts and achieve an accuracy boost up to 82%. Okumura and Muraki (1994) estimate the similarity of candidate conjuncts by means of a similarity function which incorporates syntactic, orthographical, and semantic information about the conjuncts. The model provides about 75% accuracy.",
"cite_spans": [
{
"start": 166,
"end": 192,
"text": "Agarwal and Boggess (1992)",
"ref_id": "BIBREF0"
},
{
"start": 342,
"end": 367,
"text": "Okumura and Muraki (1994)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "The resolution of coordination ambiguity can also be tried at parsing time. Charniak and Johnson (2005) , e.g., supply a discriminative reranker that uses e.g., features to capture syntactic parallelism across conjuncts. The reranker achieves an F-score of 91%.",
"cite_spans": [
{
"start": 76,
"end": 103,
"text": "Charniak and Johnson (2005)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Recently, discriminative learning-based approaches were proposed, which exploit only lexical, morpho-syntactic features and the symmetry of conjuncts. Shimbo and Hara (2007) incorporate morpho-syntactic and symmetry features in a discriminative learning model and end up with 57% F-measure on the GENIA corpus (Ohta et al., 2002) . Buyko et al. (2007) employ Conditional Random Fields (Lafferty et al., 2001 ) and successfully tested this technique in the biomedical domain for the identification and resolution of elliptified conjuncts. They evaluate on the GENIA corpus and report an F-score of 93% for the reconstruction of the elliptical conjuncts employing lexical and morpho-syntactic criteria only. At least two questions remain -whether the latter approach can achieve similar results in the newswire language domain (and is thus portable), and whether the incorporation of additional semantic criteria in this approach might boost the resolution rate, or not (and is thus possibly more parsimonious). The latter question is the main problem we deal with in this paper.",
"cite_spans": [
{
"start": 151,
"end": 173,
"text": "Shimbo and Hara (2007)",
"ref_id": "BIBREF22"
},
{
"start": 310,
"end": 329,
"text": "(Ohta et al., 2002)",
"ref_id": "BIBREF17"
},
{
"start": 332,
"end": 351,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
},
{
"start": 385,
"end": 407,
"text": "(Lafferty et al., 2001",
"ref_id": "BIBREF11"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "For our experiments, we used the WSJ part of the PENN TREEBANK (Marcus et al., 1993) . Some researchers (e.g., Hogan (2007a) ) had recently found several inconsistencies in its annotation of the bracketing of coordinations in NPs. These bugs were shown to pose problems for training and testing of coordination resolution and parsing tools. Fortunately, a re-annotated version has been provided by Vadas and Curran (2007) , with a focus on the internal structure of NPs. They added additional bracketing annotation for each noun phrase in the WSJ section of the PENN TREEBANK assuming a right-bracketing structure in NPs. In addition, they introduced tags, e.g., 'NML' for explicitly marking any left-branching constituents as in (NP (NML (JJ industrial) (CC and) (NN food)) (NNS goods)) where 'industrial' and 'food' are conjuncts. In the example (NP (DT some) (NN food) (CC and) (NN household) (NNS goods)) the structure of the noun phrase is already correct and should not be annotated further, since 'household goods' is already right-most and is coordinated with 'food'. Still, in the original PENN TREEBANK annotation, we find annotations of noun phrases such as (NP (NN royalty) (CC and) (NP (NN rock) (NNS stars))) that remain unchanged after the re-annotation process.",
"cite_spans": [
{
"start": 63,
"end": 84,
"text": "(Marcus et al., 1993)",
"ref_id": "BIBREF15"
},
{
"start": 111,
"end": 124,
"text": "Hogan (2007a)",
"ref_id": "BIBREF8"
},
{
"start": 398,
"end": 421,
"text": "Vadas and Curran (2007)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Annotation in the PENN TREEBANK",
"sec_num": "2.1"
},
{
"text": "We, first, extracted a set of 3,333 non-nested NP coordinations involving noun compounds and one conjunction, with a maximal number of nine nouns (no prepositional phrases were considered). We focused on two patterns in the re-annotated WSJ portion:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "(1) Noun phrases containing at least two nouns and a conjunction as sister nodes as in (NP (NML (NN movie) (CC and) (NN book)) (NNS pirates)) or in (NP (DT some) (NN food) (CC and) (NN household) (NNS goods)) (2) Noun phrases containing at least two noun phrases and a conjunction as sister nodes (as they remained unchanged from the original PENN TREEBANK version). Thereby, the second noun phrase contains at least two nouns as sister nodes as in (NP (NP (NNP France)) (CC and) (NP (NNP Hong) (NNP Kong)))",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "We removed from this original set NPs which could not be reduced to the following pattern: 1 1 These are typically coordinations of the form '(W )N 1 and N 2 ', e.g., 'government sources and lobbyists', where W is a sequence of i tokens (i \u2265 0). 646 coordinations of this type occurred in the WSJ portion of the PTB.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "(W ) N 1 and (W ) N 2 N 3 , where (W ) is a sequence of i tokens with i \u2265 0 as in 'street lamps N 1 and ficus N 2 trees N 3 '.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "The remaining major data set (A) then contained 2,687 NP coordinations. A second data set (B) was formed, which is a proper subset of A and contained only those coordination structures that match the following pattern:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "(X) N 1 and (W ) N 2 N 3 , where (X) is defined as a sequence of i tokens (i \u2265 0) with all part-of-speech (POS) tags except nouns and (W ) defined as above; e.g., 'a happy cat N 1 and dog N 2 owner N 3 '. Test set B contains, in our opinion, a selection of less 'hard' coordinations from the set A, and includes 1,560 items.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "All these patterns focus on three forms of conjunctions, namely 'and', 'or', and 'but not', which connect two conjuncts (the extension of which varies in our data from one up to maximally eight tokens as in 'London's \"Big Bang\" 1986 deregulation and Toronto's \"Little Bang\" the same year'.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "The remainders from the conjunctions and the two conjuncts in a coordinated NP are called shared elements (e.g., 'owner' and 'a happy' in the above example). It is evident that the correct recognition of conjunct boundaries allows for the proper identification of the shared elements.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "Set A contains 1,455 coordinations where N 1 and N 3 are coordinated (e.g, 'food and household goods') and 1,232 coordinations where N 1 and N 2 are coordinated (e.g., 'cotton and acetate fibers'). Set B consists of 643 coordinations where N 1 and N 3 are coordinated and 917 coordinations where N 1 and N 2 are coordinated.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "The extracted data sets were converted into an IO representation of tokens labeled as 'C' for conjunct, 'CC' for conjunction, and 'S' for the shared element(s). The noun phrase 'cotton and acetate fibers', e.g., is represented as a sequence 'C CC C S', while 'food and household goods' is represented as a sequence 'C CC C C'.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Coordination Corpus",
"sec_num": "2.2"
},
{
"text": "We here compare three different approaches to the resolution of noun phrase coordination ambiguity, viz. ones relying solely on morpho-syntactic information, solely on lexico-semantic information, and a cumulative combination of both. As far as semantic information is concerned we make use of various WORDNET similarity measures.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Methods",
"sec_num": "3"
},
{
"text": "We used three baselines for resolving noun phrase coordination ambiguities -one incorporating only lexico-semantic information, the WordNet Similarity baseline, and two alternative ones incorporating only morpho-syntactic and syntactic parse information, the Number Agreement and the Bikel Parser baseline, respectively.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Baselines",
"sec_num": "3.1"
},
{
"text": "Our lexico-semantic baseline comes with WORDNET semantic similarity scores of putatively coordinated nouns. For our experiments, we used the implementation of WORDNET similarity and relatedness measures provided by Ted Pedersen. 2 The following similarity measures were considered: two measures based on path lenghts between concepts (path and lch (Leacock et al., 1998) ), three measures based on information content, i.e., corpus-based measures of the specificity of a concept (res (Resnik, 1999) , lin (Lin, 1998) , and jcn (Jiang and Conrath, 1997) ). Furthermore, we used two relatedness measures, namely, lesk (Banerjee and Pedersen, 2003) and vector (Patwardhan et al., 2003) , which score the similarity of the glosses of both concepts. We applied these similarity measures to any pair of putatively coordinated nouns in the noun phrases from our data sets, A and B. To determine potential conjuncts we calculate two similarity scores relative to the structures discussed in Section 2.2:",
"cite_spans": [
{
"start": 229,
"end": 230,
"text": "2",
"ref_id": null
},
{
"start": 348,
"end": 370,
"text": "(Leacock et al., 1998)",
"ref_id": "BIBREF12"
},
{
"start": 484,
"end": 498,
"text": "(Resnik, 1999)",
"ref_id": "BIBREF20"
},
{
"start": 505,
"end": 516,
"text": "(Lin, 1998)",
"ref_id": "BIBREF13"
},
{
"start": 527,
"end": 552,
"text": "(Jiang and Conrath, 1997)",
"ref_id": "BIBREF10"
},
{
"start": 616,
"end": 645,
"text": "(Banerjee and Pedersen, 2003)",
"ref_id": "BIBREF1"
},
{
"start": 657,
"end": 682,
"text": "(Patwardhan et al., 2003)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "WORDNET Similarity (WN) Baseline",
"sec_num": "3.1.1"
},
{
"text": "s 1 = sim(N 1 , N 2 ) and s 2 = sim(N 1 , N 3 )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "WORDNET Similarity (WN) Baseline",
"sec_num": "3.1.1"
},
{
"text": "Our final score is the maximum over both scores which is then the semantic indicator for the most plausible resolution of the coordination.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "WORDNET Similarity (WN) Baseline",
"sec_num": "3.1.1"
},
{
"text": "We compared here the number agreement between selected nouns (see Resnik (1999) ",
"cite_spans": [
{
"start": 66,
"end": 79,
"text": "Resnik (1999)",
"ref_id": "BIBREF20"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Number Agreement (NA) Baseline",
"sec_num": "3.1.2"
},
{
"text": "). Accordingly, N 1 and N 2 are coordinated, if number(N 1 ) = number(N 2 ) AND number(N 1 ) = number(N 3 ), while N 1 and N 3 are coordinated, if number(N 1 ) = number(N 3 ) AND number(N 1 ) = number(N 2 ).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Number Agreement (NA) Baseline",
"sec_num": "3.1.2"
},
{
"text": "In the WN and NA baselines, after the detection of coordinated elements we used simple heuristics to tag the remaining part of the noun phrase. If N 1 and N 2 were hypothesized to be coordinated, then all tokens preceding N 1 were tagged as shared elements, N 3 was tagged as shared element as well, while all tokens between the conjunction and N 2 were tagged as conjuncts. For example, in 'a happy dog N 1 and cat N 2 owner N 3 ' we identify 'dog' and 'cat' as coordinated elements and tag 'a happy' and 'owner' as shared elements. The final resolution looks like 'S S C CC C S'. If N 1 and N 3 were hypothesized to be coordinated, then all other elements except conjunctions were tagged as parts of conjuncts, as well.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Post-Processing Heuristics",
"sec_num": "3.1.3"
},
{
"text": "We used the well-known Bikel Parser (Bikel, 2004) in its original version and the one used by Collins (2003) . We trained both of them only with NPs extracted from the re-annotated version of WSJ (see Section 2) and converted the bracketing output of the parsers to the IO representation for NP coordinations for further evaluations.",
"cite_spans": [
{
"start": 36,
"end": 49,
"text": "(Bikel, 2004)",
"ref_id": "BIBREF2"
},
{
"start": 94,
"end": 108,
"text": "Collins (2003)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Bikel Parser (BP) Baseline",
"sec_num": "3.1.4"
},
{
"text": "The approach to conjunct identification presented by Buyko et al. (2007) employs Conditional Random Fields (CRF) (Lafferty et al., 2001 ), 3 which assign a label to each token of coordinated NPs according to its function in the coordination: 'C' for conjuncts, 'CC' for conjunctions, and 'S' for shared elements. Since non-nested conjuncts can be assumed to be in a sequential order, sequential learning approaches (instead of single position classification approaches) seem appropriate here. Buyko et al. (2007) report an F-measure of 93% on conjunct identification in the GENIA corpus. They use a feature set including lexical (words), and morpho-syntactic features (POS tags, morphosyntactic similarity of putative conjuncts), but exclude any semantic criteria. The morpho-syntactic similarity features were generated from a rulebased approach to conjunct identification using the maximal symmetry of conjuncts as constituted by their respective POS annotation.",
"cite_spans": [
{
"start": 53,
"end": 72,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
},
{
"start": 113,
"end": 135,
"text": "(Lafferty et al., 2001",
"ref_id": "BIBREF11"
},
{
"start": 493,
"end": 512,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Chunking of Conjuncts with CRFs",
"sec_num": "3.2"
},
{
"text": "We here intend to apply this approach for resolving coordination ambiguities involving noun compounds in the newswire language such as 'president and chief executive'. This restricts the spectrum of considered coordinations in noun phrases to more complicated cases than those considered by Buyko et al. (2007) . We will thus test the various resolution models under harder test conditions, Feature Class Description default feature prior probability distribution over the possible argument labels lexical word morphosyntactic the token's POS tag; output labels of the morpho-syntactic similarity ('C','CC' and 'S') (see Buyko et al. (2007) ); output labels of the number agreement baseline ('C', 'CC' and 'S') semantic WN output labels of the WORDNET similarity baseline ('C', 'CC' and 'S') contextual conjunctions of all features of neighboring tokens (two tokens to the left and one token to the right) We also propose in this work an extension of the feature space in terms of lexico-semantic features (see Table 1 ), information that originates from similarity computations on WORDNET data. Furthermore, we do not use orthographical features of the original approach as they are well suited only for the biomedical language domain.",
"cite_spans": [
{
"start": 291,
"end": 310,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
},
{
"start": 621,
"end": 640,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [
{
"start": 1011,
"end": 1018,
"text": "Table 1",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Chunking of Conjuncts with CRFs",
"sec_num": "3.2"
},
{
"text": "To evaluate the different approaches to conjunct identification, we used recall and precision scores since they are well suited for the evaluation of segmentation tasks. Two types of decisions were evaluated -the assignment of 'C' labels denoting conjuncts in terms of the F-measure, and (given the tagged conjuncts) the accuracy of the complete coordination resolution. A coordination is resolved properly only, if all tokens of both conjuncts are correctly identified.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "We carried out a ten-fold cross-validation of all ML-based methods (Bikel parser (Bikel, 2004) and CRF-based conjunct identification (Buyko et al., 2007) ). For the evaluation of the NA and WN baselines, we tested their performance on the complete data sets, A and B (see Section 2).",
"cite_spans": [
{
"start": 67,
"end": 94,
"text": "(Bikel parser (Bikel, 2004)",
"ref_id": null
},
{
"start": 133,
"end": 153,
"text": "(Buyko et al., 2007)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "As Table 2 depicts the NA baseline achieved an accuracy of 28.4% on A (36.6% on B), the Bikel parser reached 77.2% on A (73.4% in B), while the WN baseline got in its best run (vector measure) an accuracy of 41.7% on A (49.6% on B). These results already reveal that parsing almost dramatically outperforms the coordination resolution based on the NA similarity by up to 35.5% points. The results of the WN baseline indicate that the best similarity measure for conjunct identification is the vector similarity (Patwardhan et al., 2003) that scores the similarity between the glosses of the concepts.",
"cite_spans": [
{
"start": 511,
"end": 536,
"text": "(Patwardhan et al., 2003)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [
{
"start": 3,
"end": 10,
"text": "Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "Our error analysis of the WN baseline on the test set A reveals that its low accuracy has various reasons. First, about 37% of the coordinations could not be resolved due to the absence of at least one noun involved in the coordination from the WORDNET. These coordinations usually include named entities such as person and organization names (e.g., 'brothers Francis and Gilbert Gros'). These coverage gaps have clearly a negative effect on the resolution results for all WORD-NET similarity measures.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "To find out errors which are specific for the considered similarity measures, we have chosen the res measure and inspected the analysis results on all noun phrases where nouns are covered by WORDNET. The remaining set of coordinations contains 1,740 noun phrases. 1,022 coordinations (59%) of this set were completely resolved by the WN baseline, while 1,117 coordinations (64% of the remaining part, 41.5% of the test set A) could be at least partly resolved. Obviously, the coordinated heads are properly detected by the res measure but our heuristics for tagging the remaining modifiers (see Subsection 3.1.3) fail to provide the correct conjunct boundaries.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "623 coordinations (36%) were mis-classified by the res measure. A closer look at this data reveals two types of errors. The first and minor type is the misleading selection of putatively coordinated heads N 1 , N 2 , and N 3 . We presuppose in the WN baseline that the heads appear rightmost in the noun phrase, although that is not always the case as illustrated by the phrase 'Northern California earthquake and Hurricane Hugo'. The res measure detected correctly a higher similarity between 'earthquake' (N 1 ) and 'hurricane' (N 2 ), but 'Hugo' (N 3 ) is a modifier of 'hurricane'. Although the res measure works fine, the coordination cannot be properly resolved due to syntactic reasons. In some cases, N 2 is wrongly selected as in 'life and health insurance operation' where the WN baseline selects 'insurance' as right-most noun (except the last noun 'operation') and not 'health '. 4 Set A Set B Recall/Precision/F-Score Accuracy Recall/Precision/F-Score Accuracy NA 32.7 / 75.9 / 45.7 28.4 41.6 / 83.9 / 55.6 36. The second type of error comes as erroneous classifications of the res measure such as in 'hospitals and blood banks' where 'hospitals' and 'blood' have a higher similarity than 'hospitals' and 'banks' although they are, in fact, coordinated here. 'hotels and large restaurant chains', 'records and music publishing', 'chemicals and textiles company' are other examples for the observation that the coordinated elements have a lower similarity as non-coordinated ones.",
"cite_spans": [
{
"start": 889,
"end": 893,
"text": "'. 4",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "We also carried out a ten-fold cross-validation of the CRF-based approach for the conjunct identification. First of all, the CRF-based approach (with and without WN similarity) achieved the highest accuracy score -up to 78.0% on set A, and 79.1% on B -compared with all other approaches we scrutinized on. We also tested the performance of the original semantics-free approach and the additional effects of the WORDNET similarity measures (see Table 2 ). Although the integration of semantic information leads to a mild gain compared with the original approach (up to 0.3% points, with the lesk measure), the results indicate that no substantial benefit can be traced to semantic features.",
"cite_spans": [],
"ref_spans": [
{
"start": 444,
"end": 451,
"text": "Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "We ran several tests with solely morphosyntactic features (as enumerated in Table 1) and solely WN features, too. They reveal that solely morpho-syntactic features are up to 11.8% points more predictive than WN features. The best results were still achieved using the gloss-oriented lesk measure (see Table 3 ).",
"cite_spans": [],
"ref_spans": [
{
"start": 76,
"end": 84,
"text": "Table 1)",
"ref_id": "TABREF0"
},
{
"start": 301,
"end": 308,
"text": "Table 3",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "The inspection of the errors types from the various runs is not fully conclusive though. After adding WN features to both sets, we detected some improvements for conjunct tagging with high WN similarity. Some conjunct boundaries could be corrected as in 'record and movie producer' where, in the first run, 'producer' was tagged as a conjunct and was corrected as being shared by integrating WN features. But we also detected a degrading tagging behavior of conjuncts with WN features where the WN similarity was not helpful at all as in 'chairman and chief designer' where 'chairman' and 'chief' under the influence of WN features were judged to be conjuncts. We found out that the addition of WN features positively influences the classification of coordinations where N 1 and N 2 are coordinated, while it increased errors in the classification of coordinations where N 1 and N 3 are coordinated.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "In addition, we calculated intersections between the set A error data (unique) of the res WN baseline (1400 phrases) and the error data of the CRF approach without WN features (391), and the error data of the CRF approach with WN features (385), respectively. These error data sets contain noun phrases where coordinated heads could not be properly detected. The set of the res WN baseline and the set of the CRF approach without WN features have an intersection of 230 instances, where 138 instances could not be found in the WORDNET. That means that for about 161 instances (59%) in the mis-classified data of the CRF approach the additional WN features would not be helpful. The intersection remains similar (226 instances) between the set of the res WN base- line and the set of the CRF approach enriched with WN features. The intersection between the errror sets of the both CRF approaches includes 352 instances. The integration of the WN features was not helful for almost the complete error data from the original CRF approach. We have previously shown that the res WN baseline features correlate with the correct label sequence for only 1,117 coordinations (41.5%) of the complete evaluation set A and the features thus do not seem to be effective in our approach. Furthermore, we evaluated the results of the CRF approach only for the correct detection of coordinated heads (see above for the res measure and intersection counts) and disregarded the modifier classification. The results -85.3% on set A and 85.4% on set B -reveal that the classification of modifiers is a major source of classification errors. In both configurations the problematic noun phrases are the ones with (e.g., adjectival) modifiers. The boundaries of conjuncts are not properly recognized in such noun phrases, as for example in 'American comic book and television series' where the correct label sequence is 'S C C CC C S', since 'American' is the shared modifier of 'book ' and 'television', while 'comic' just modifies 'book'. As most adjectives appearing at the beginning of the noun phrase as in 'medical products and services company' tend to be used as shared modifiers of coordinations in our data, this, erroneously, leads to false taggings, e.g., 'personal' in 'personal computer and software design' as a shared element. To cope adequately with modifiers we need to integrate more appropriate features such as collocation frequencies of modifiers and coordinated heads. The detection of a higher collocation frequency of 'personal computer' in comparison to 'personal software' (e.g., using the pro-cedures proposed by Wermter and Hahn (2004) ) would help tagging the conjunct boundaries.",
"cite_spans": [
{
"start": 1962,
"end": 2017,
"text": "' and 'television', while 'comic' just modifies 'book'.",
"ref_id": null
},
{
"start": 2618,
"end": 2641,
"text": "Wermter and Hahn (2004)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Results and Error Analysis",
"sec_num": "4"
},
{
"text": "We investigated the problem of noun phrase coordination resolution as a segmentation problem among conjuncts involved in the coordination. While resolving coordination ellipsis is often considered as a semantically constrained problem, we wanted to assess a less 'costly' solution strategy, namely relying on 'cheaper' to get syntactic criteria as much as possible, though not sacrificing the accurary of resolutions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Work",
"sec_num": "5"
},
{
"text": "We, first looked at morpho-syntactic criteria only and lexico-semantic criteria only, and then at the combination of both approaches. The evaluation results from a variety of experiments reveal that the major part of ambiguous coordinations can be resolved using solely morpho-syntactic features. Surprising as it might be, the semantic information as derived from the WORDNET similarity measures does not yield any further substantial improvement for our approach. This is somehow counter-intuitive, but our findings, unlike those from earlier studies which emphasized the role of semantic criteria, are based on extensive corpus data -the PENN TREEBANK.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Work",
"sec_num": "5"
},
{
"text": "Results from our error analysis will guide future work to further boost results. Particular emphasis will be laid on the integration of named entity recognizers, collocation frequencies and distributional similarity data as also advocated by Chantree et al. (2005) .",
"cite_spans": [
{
"start": 242,
"end": 264,
"text": "Chantree et al. (2005)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Work",
"sec_num": "5"
},
{
"text": "The presented sequential labeling-based approach to coordination resolution was here applied to the resolution of a special type of ambiguous noun phrases. In general, this approach can easily be applied to the resolution of other types of coordinative structures in noun phrases as already presented in Buyko et al. (2007) . As far as other phrasal types (e.g., verbal phrases) are concerned, long-distance coordinations play a much more prominent role. The token-based labeling approach may be thus substituted by a chunk-based approach operating on sentences.",
"cite_spans": [
{
"start": 304,
"end": 323,
"text": "Buyko et al. (2007)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Work",
"sec_num": "5"
},
{
"text": "http://www.d.umn.edu/\u223ctpederse/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "They employ the linear-chain CRF implementation from the MALLET toolsuite available at http://mallet.cs. umass.edu/index.php/Main_Page",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "'turbine N 2 ' is, however, correctly selected in 'steam turbine and gas turbine plants'.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "This research was partly funded by the German Ministry of Education and Research within the STEMNET project (01DS001A-C) and by the EC within the BOOTSTREP project (FP6-028099).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "A simple but useful approach to conjunct identification",
"authors": [
{
"first": "R",
"middle": [],
"last": "Agarwal",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Boggess",
"suffix": ""
}
],
"year": 1992,
"venue": "Proceedings of the 30th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "15--21",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Agarwal, R. and L. Boggess. 1992. A simple but useful approach to conjunct identification. In Pro- ceedings of the 30th Annual Meeting of the Asso- ciation for Computational Linguistics, pages 15-21. Newark, DE, USA, 28 June -2 July 1992.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Extended gloss overlaps as a measure of semantic relatedness",
"authors": [
{
"first": "S",
"middle": [],
"last": "Banerjee",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Pedersen",
"suffix": ""
}
],
"year": 2003,
"venue": "IJ-CAI'03 -Proceedings of the 18th International Joint Conference on Artificial Intelligence",
"volume": "",
"issue": "",
"pages": "805--810",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Banerjee, S. and T. Pedersen. 2003. Extended gloss overlaps as a measure of semantic relatedness. In IJ- CAI'03 -Proceedings of the 18th International Joint Conference on Artificial Intelligence, pages 805- 810. Acapulco, Mexico, August 9-15, 2003.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Intricacies of Collins' parsing model",
"authors": [
{
"first": "D",
"middle": [],
"last": "Bikel",
"suffix": ""
}
],
"year": 2004,
"venue": "Computational Linguistics",
"volume": "30",
"issue": "4",
"pages": "479--511",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bikel, D. 2004. Intricacies of Collins' parsing model. Computational Linguistics, 30(4):479-511.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Resolution of coordination ellipses in biological named entities using Conditional Random Fields",
"authors": [
{
"first": "E",
"middle": [],
"last": "Buyko",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Tomanek",
"suffix": ""
},
{
"first": "U",
"middle": [],
"last": "Hahn",
"suffix": ""
}
],
"year": 2007,
"venue": "PAC-LING 2007 -Proceedings of the 10th Conference of the Pacific Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "163--171",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Buyko, E., K. Tomanek, and U. Hahn. 2007. Reso- lution of coordination ellipses in biological named entities using Conditional Random Fields. In PAC- LING 2007 -Proceedings of the 10th Conference of the Pacific Association for Computational Linguis- tics, pages 163-171. Melbourne, Australia, Septem- ber 19-21, 2007.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Disambiguating coordinations using word distribution information",
"authors": [
{
"first": "F",
"middle": [
"A"
],
"last": "Chantree",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Kilgarriff",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "De Roeck",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Willis",
"suffix": ""
}
],
"year": 2005,
"venue": "RANLP 2005 -Proceedings of the Intl. Conference on 'Recent Advances in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "21--23",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chantree, F. A. Kilgarriff, A. de Roeck, and A. Willis. 2005. Disambiguating coordinations using word dis- tribution information. In RANLP 2005 -Proceed- ings of the Intl. Conference on 'Recent Advances in Natural Language Processing', pages 144-151. Borovets, Bulgaria, 21-23 September, 2005.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Coarse-to-fine n-best parsing and MaxEnt discriminative reranking",
"authors": [
{
"first": "E",
"middle": [],
"last": "Charniak",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Johnson",
"suffix": ""
}
],
"year": 2005,
"venue": "ACL'05 -Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "25--30",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Charniak, E. and M. Johnson. 2005. Coarse-to-fine n-best parsing and MaxEnt discriminative reranking. In ACL'05 -Proceedings of the 43rd Annual Meet- ing of the Association for Computational Linguistics, pages 173-180. Ann Arbor, MI, 25-30 June 2005.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Head-driven statistical models for natural language parsing",
"authors": [
{
"first": "M",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2003,
"venue": "Computational Linguistics",
"volume": "29",
"issue": "4",
"pages": "589--637",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Collins, M. 2003. Head-driven statistical models for natural language parsing. Computational Linguis- tics, 29(4):589-637.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "WORDNET: An Electronic Lexical Database",
"authors": [
{
"first": "C",
"middle": [],
"last": "Fellbaum",
"suffix": ""
}
],
"year": 1998,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Fellbaum, C., editor. 1998. WORDNET: An Electronic Lexical Database. MIT Press.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Coordinate noun phrase disambiguation in a generative parsing model",
"authors": [
{
"first": "D",
"middle": [],
"last": "Hogan",
"suffix": ""
}
],
"year": 2007,
"venue": "ACL'07 -Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics",
"volume": "",
"issue": "",
"pages": "680--687",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hogan, D. 2007a. Coordinate noun phrase disam- biguation in a generative parsing model. In ACL'07 -Proceedings of the 45th Annual Meeting of the As- sociation of Computational Linguistics, pages 680- 687. Prague, Czech Republic, June 28-29, 2007.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Empirical measurements of lexical similarity in noun phrase conjuncts",
"authors": [
{
"first": "D",
"middle": [],
"last": "Hogan",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics. Demo and Poster Sessions",
"volume": "",
"issue": "",
"pages": "149--152",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hogan, D. 2007b. Empirical measurements of lexi- cal similarity in noun phrase conjuncts. In Proceed- ings of the 45th Annual Meeting of the Association of Computational Linguistics. Demo and Poster Ses- sions, pages 149-152. Prague, Czech Republic, June 28-29, 2007.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Semantic similarity based on corpus statistics and lexical taxonomy",
"authors": [
{
"first": "J",
"middle": [],
"last": "Jiang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Conrath",
"suffix": ""
}
],
"year": 1997,
"venue": "ROCLING-X -Proceedings of the 1997 International Conference on Research in Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jiang, J. and D. Conrath. 1997. Semantic similar- ity based on corpus statistics and lexical taxonomy. In ROCLING-X -Proceedings of the 1997 Inter- national Conference on Research in Computational Linguistics. Taipei, Taiwan, August 22-24, 1997.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Conditional Random Fields: Probabilistic models for segmenting and labeling sequence data",
"authors": [
{
"first": "J",
"middle": [],
"last": "Lafferty",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Mccallum",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Pereira",
"suffix": ""
}
],
"year": 2001,
"venue": "ICML-2001 -Proceedings of the 18th International Conference on Machine Learning",
"volume": "",
"issue": "",
"pages": "282--289",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lafferty, J., A. McCallum, and F. Pereira. 2001. Condi- tional Random Fields: Probabilistic models for seg- menting and labeling sequence data. In ICML-2001 -Proceedings of the 18th International Conference on Machine Learning, pages 282-289. Williams College, MA, USA, June 28 -July 1, 2001.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Using corpus statistics and WORDNET relations for sense identification",
"authors": [
{
"first": "C",
"middle": [],
"last": "Leacock",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Chodorow",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Miller",
"suffix": ""
}
],
"year": 1998,
"venue": "Computational Linguistics",
"volume": "24",
"issue": "1",
"pages": "147--165",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Leacock, C., M. Chodorow, and G. Miller. 1998. Using corpus statistics and WORDNET relations for sense identification. Computational Linguistics, 24(1):147-165.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "An information-theoretic definition of similarity",
"authors": [
{
"first": "D",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 1998,
"venue": "Proceedings of the 15th International Conference on Machine Learning",
"volume": "",
"issue": "",
"pages": "296--304",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lin, D. 1998. An information-theoretic definition of similarity. In Proceedings of the 15th International Conference on Machine Learning, pages 296-304.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Building a large annotated corpus of English: The PENN TREEBANK",
"authors": [
{
"first": "M",
"middle": [],
"last": "Marcus",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Santorini",
"suffix": ""
},
{
"first": "M.-A",
"middle": [],
"last": "Marcinkiewicz",
"suffix": ""
}
],
"year": 1993,
"venue": "Computational Linguistics",
"volume": "19",
"issue": "2",
"pages": "313--330",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Marcus, M., B. Santorini, and M.-A. Marcinkiewicz. 1993. Building a large annotated corpus of English: The PENN TREEBANK. Computational Linguistics, 19(2):313-330.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Using the Web as an implicit training set: Application to structural ambiguity resolution",
"authors": [
{
"first": "P",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Hearst",
"suffix": ""
}
],
"year": 2005,
"venue": "HLT-EMNLP'05 -Proceedings of the 5th Human Language Technology Conference and 2005 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "835--842",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nakov, P. and M. Hearst. 2005. Using the Web as an implicit training set: Application to structural ambi- guity resolution. In HLT-EMNLP'05 -Proceedings of the 5th Human Language Technology Conference and 2005 Conference on Empirical Methods in Nat- ural Language Processing, pages 835-842. Vancou- ver, B.C., Canada, October 6-8, 2005.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "The GE-NIA corpus: An annotated research abstract corpus in molecular biology domain",
"authors": [
{
"first": "T",
"middle": [],
"last": "Ohta",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Tateisi",
"suffix": ""
},
{
"first": "J.-D",
"middle": [],
"last": "Kim",
"suffix": ""
}
],
"year": 2002,
"venue": "HLT 2002 -Proceedings of the 2nd International Conference on Human Language Technology Research",
"volume": "",
"issue": "",
"pages": "82--86",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ohta, T., Y. Tateisi, and J.-D. Kim. 2002. The GE- NIA corpus: An annotated research abstract corpus in molecular biology domain. In HLT 2002 -Pro- ceedings of the 2nd International Conference on Hu- man Language Technology Research, pages 82-86. San Diego, CA, USA, March 24-27, 2002.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Symmetric pattern matching analysis for English coordinate structures",
"authors": [
{
"first": "A",
"middle": [],
"last": "Okumura",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Muraki",
"suffix": ""
}
],
"year": 1994,
"venue": "ANLP 1994 -Proceedings of the 4th Conference on Applied Natural Language Processing",
"volume": "",
"issue": "",
"pages": "13--15",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Okumura, A. and K. Muraki. 1994. Symmetric pattern matching analysis for English coordinate structures. In ANLP 1994 -Proceedings of the 4th Conference on Applied Natural Language Processing, pages 41- 46. Stuttgart, Germany, 13-15 October 1994.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Using measures of semantic relatedness for word sense disambiguation",
"authors": [
{
"first": "S",
"middle": [],
"last": "Patwardhan",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Banerjee",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Pedersen",
"suffix": ""
}
],
"year": 2003,
"venue": "CICLing 2003 -Proceedings 4th Intl. Conference on Computational Linguistics and Intelligent Text Processing",
"volume": "",
"issue": "",
"pages": "241--257",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Patwardhan, S., S. Banerjee, and T. Pedersen. 2003. Using measures of semantic relatedness for word sense disambiguation. In CICLing 2003 -Proceed- ings 4th Intl. Conference on Computational Linguis- tics and Intelligent Text Processing, pages 241-257. Mexico City, Mexico, February 16-22, 2003.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Semantic similarity in a taxonomy: An information-based measure and its application to problems of ambiguity in natural language",
"authors": [
{
"first": "P",
"middle": [],
"last": "Resnik",
"suffix": ""
}
],
"year": 1999,
"venue": "Journal of Artificial Intelligence Research",
"volume": "11",
"issue": "",
"pages": "95--130",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Resnik, P. 1999. Semantic similarity in a taxonomy: An information-based measure and its application to problems of ambiguity in natural language. Journal of Artificial Intelligence Research, 11:95-130.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Bracketing compound nouns for logic form derivation",
"authors": [
{
"first": "V",
"middle": [],
"last": "Rus",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Moldovan",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Bolohan",
"suffix": ""
}
],
"year": 2002,
"venue": "FLAIRS 2002 -Proceedings of the 15th International Florida Artificial Intelligence Research Society Conference",
"volume": "",
"issue": "",
"pages": "198--202",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rus, V., D. Moldovan, and O. Bolohan. 2002. Bracket- ing compound nouns for logic form derivation. In FLAIRS 2002 -Proceedings of the 15th Interna- tional Florida Artificial Intelligence Research Soci- ety Conference, pages 198-202. Pensacola Beach, FL, USA, May 14-16, 2002.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "A discriminative learning model for coordinate conjunctions",
"authors": [
{
"first": "M",
"middle": [],
"last": "Shimbo",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Hara",
"suffix": ""
}
],
"year": 2007,
"venue": "EMNLP-CoNLL 2007 -Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning",
"volume": "",
"issue": "",
"pages": "610--619",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Shimbo, M. and K. Hara. 2007. A discriminative learn- ing model for coordinate conjunctions. In EMNLP- CoNLL 2007 -Proceedings of the 2007 Joint Con- ference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 610-619. Prague,Czech Republic, June 28-29, 2007.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Adding noun phrase structure to the PENN TREEBANK",
"authors": [
{
"first": "D",
"middle": [],
"last": "Vadas",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Curran",
"suffix": ""
}
],
"year": 2007,
"venue": "ACL'07 -Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics",
"volume": "",
"issue": "",
"pages": "240--247",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Vadas, D. and J. Curran. 2007. Adding noun phrase structure to the PENN TREEBANK. In ACL'07 -Pro- ceedings of the 45th Annual Meeting of the Associ- ation of Computational Linguistics, pages 240-247. Prague, Czech Republic, June 28-29, 2007.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Collocation extraction based on modifiability statistics",
"authors": [
{
"first": "J",
"middle": [],
"last": "Wermter",
"suffix": ""
},
{
"first": "U",
"middle": [],
"last": "Hahn",
"suffix": ""
}
],
"year": 2004,
"venue": "COLING 2004 -Proceedings of the 20th International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "980--986",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wermter, J. and U. Hahn. 2004. Collocation extraction based on modifiability statistics. In COLING 2004 - Proceedings of the 20th International Conference on Computational Linguistics, pages 980-986. Geneva, Switzerland, August 23-27, 2004.",
"links": null
}
},
"ref_entries": {
"TABREF0": {
"html": null,
"num": null,
"text": "",
"type_str": "table",
"content": "<table><tr><td>: Feature Classes Used for Conjunct Iden-</td></tr><tr><td>tification</td></tr><tr><td>since Buyko et al. consider, e.g., adjective coordi-</td></tr><tr><td>nations in noun phrases such as 'positive and neg-</td></tr><tr><td>ative IL-2 regulator' that are predominant in the</td></tr><tr><td>biomedical language domain.</td></tr></table>"
},
"TABREF2": {
"html": null,
"num": null,
"text": "",
"type_str": "table",
"content": "<table/>"
},
"TABREF4": {
"html": null,
"num": null,
"text": "Conjunct Identification -Crossvalidation on the WSJ section of the PENN TREE-",
"type_str": "table",
"content": "<table/>"
}
}
}
} |