File size: 59,654 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 | {
"paper_id": "2020",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T13:05:04.825196Z"
},
"title": "A study of semantic projection from single word terms to multi-word terms in the environment domain",
"authors": [
{
"first": "Yizhe",
"middle": [],
"last": "Wang",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "CNRS & University of Nantes",
"location": {
"postCode": "LN2S"
}
},
"email": "yizhe.wang@univ-tlse2.fr"
},
{
"first": "B\u00e9atrice",
"middle": [],
"last": "Daille",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "CNRS & University of Nantes",
"location": {
"postCode": "LN2S"
}
},
"email": "beatrice.daille@univ-nantes.fr"
},
{
"first": "Nabil",
"middle": [],
"last": "Hathout",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "CNRS & University of Nantes",
"location": {
"postCode": "LN2S"
}
},
"email": "nabil.hathout@univ-tlse2.fr"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "The semantic projection method is often used in terminology structuring to infer semantic relations between terms. Semantic projection relies upon the assumption of semantic compositionality: the relation that links simple term pairs remains valid in pairs of complex terms built from these simple terms. This paper proposes to investigate whether this assumption commonly adopted in natural language processing is actually valid. First, we describe the process of constructing a list of semantically linked multi-word terms (MWTs) related to the environmental field through the extraction of semantic variants. Second, we present our analysis of the results from the semantic projection. We find that contexts play an essential role in defining the relations between MWTs.",
"pdf_parse": {
"paper_id": "2020",
"_pdf_hash": "",
"abstract": [
{
"text": "The semantic projection method is often used in terminology structuring to infer semantic relations between terms. Semantic projection relies upon the assumption of semantic compositionality: the relation that links simple term pairs remains valid in pairs of complex terms built from these simple terms. This paper proposes to investigate whether this assumption commonly adopted in natural language processing is actually valid. First, we describe the process of constructing a list of semantically linked multi-word terms (MWTs) related to the environmental field through the extraction of semantic variants. Second, we present our analysis of the results from the semantic projection. We find that contexts play an essential role in defining the relations between MWTs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Terminology is structured by semantic relations between terms. The relations may be identified by experts, obtained from existing resources or extracted from corpora. They include synonymy, quasi-synonymy, hypernymy, hyponymy, etc. In this study, we focus on the identification of terminological relations between multi-word terms (MWTs) using the semantic projection method. We built a set of French MWT candidates related to the environment domain and containing two lexical words such as r\u00e9chauffement climatique 'global warming'. Three relation categories (antonymy, quasi-synonym, and hypernymy) between single word terms (SWTs) are extended to these candidates. A subset of these MWT pairs has been validated by three judges to assess the preservation and the validity of the inferred relations between MWTs. The main finding of the evaluation is that the context is crucial for the assessment because they determine the actual meaning of the MWTs. The remaining of the paper is organized as follows. Section 2. presents the related work. Section 3. outlines the projection of the semantic relations on the MWT pairs. Section 4. describes the data and resources used for the generation of semantically linked MWTs (Section 5.). The manual evaluation and an analysis of the projection results are presented in Section 6.. A short conclusion is then given in Section 7..",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "Several approaches to semantic relation recognition have been proposed in the literature. They may be classified into three types: lexicon-based approaches (Senellart and Blondel, 2008) ; pattern-based approaches (Wang et al., 2010; Ravichandran and Hovy, 2002) ; distributional approaches (Rajana et al., 2017; Shwartz and Dagan, 2019) . Since MWTs are compositional or at least weakly compositional (L'homme, 2004) , the semantic projection method, also known as semantic variation and often referred to as a compositional method, is widely used to generate MWTs and predict relations between them from semantically related SWTs.",
"cite_spans": [
{
"start": 156,
"end": 185,
"text": "(Senellart and Blondel, 2008)",
"ref_id": "BIBREF13"
},
{
"start": 213,
"end": 232,
"text": "(Wang et al., 2010;",
"ref_id": "BIBREF15"
},
{
"start": 233,
"end": 261,
"text": "Ravichandran and Hovy, 2002)",
"ref_id": "BIBREF11"
},
{
"start": 290,
"end": 311,
"text": "(Rajana et al., 2017;",
"ref_id": "BIBREF10"
},
{
"start": 312,
"end": 336,
"text": "Shwartz and Dagan, 2019)",
"ref_id": "BIBREF14"
},
{
"start": 401,
"end": 416,
"text": "(L'homme, 2004)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2."
},
{
"text": "Synonymy is an important relation in terminology and is addressed in several studies, like Hamon et al. (1998) who identify synonymous term candidates through inference rules. The authors extract MWT candidates from a corpus on electric power plants and analyze the candidate terms (ligne d'alimentation 'supply line') as being made up of a head (ligne 'line') and an expansion (alimentation 'supply'). They then replace the head or the expansion (or both) by their synonyms obtained from a general dictionary. They assume that the substitution preserves the synonymy relation. In their study, 396 MWT pairs have been validated by an expert; 37% are real synonyms. The same method is used by Hamon and Nazarenko (2001) in order to detect synonymous MWTs in specialized corpora on nuclear power plants and coronary diseases. Their results show that general language dictionaries complement specialized hand-built lexical resources for the detection of semantic variants.",
"cite_spans": [
{
"start": 91,
"end": 110,
"text": "Hamon et al. (1998)",
"ref_id": "BIBREF5"
},
{
"start": 692,
"end": 718,
"text": "Hamon and Nazarenko (2001)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2."
},
{
"text": "In a similar study, Morin (1999) uses inference rules to identify hierarchical relations (hypernymy) between MWTs. Instead of using relations from the general dictionary, they take as reference semantically linked SWTs extracted from the AGROVOC terminology. They not only add syntactic and semantic constraints on the reference rules but also use the semantic relations with morphological relations to detect the semantic variants. They then compare the relations generated from AGROVOC with relations generated from a general language dictionary and show that the latter has a significantly lower precision. More recently, Daille and Hazem (2014) have generalized the projection method to all types of lexical relations while Hazem and Daille (2018) use it to extract synonymous MWTs with variable lengths.",
"cite_spans": [
{
"start": 20,
"end": 32,
"text": "Morin (1999)",
"ref_id": "BIBREF9"
},
{
"start": 625,
"end": 648,
"text": "Daille and Hazem (2014)",
"ref_id": "BIBREF2"
},
{
"start": 728,
"end": 751,
"text": "Hazem and Daille (2018)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2."
},
{
"text": "The main difference between our study and the ones presented above is that we use the context to validate the inferred relations. In our experiment, we have extracted from the corpus 5 contexts for each candidate in the validation dataset. We consider that the projection is valid if the meaning of two MTWs in at least two of their contexts is in the relation stated between the two SWTs that yielded them. The above studies do not use the context except (Hamon and Nazarenko, 2001 ) who checks whether the two MWT candidates can be substituted one for the other in one context. In other words, one contribution of our study is to take into account the possible ambiguity of the MWTs, and the way contexts determine their meanings.",
"cite_spans": [
{
"start": 456,
"end": 482,
"text": "(Hamon and Nazarenko, 2001",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2."
},
{
"text": "Our method is based on the assumption that MWT meaning is compositional. One consequence of this hypothesis is that when two MWTs t 1 and t 2 only differ by one of their components c 1 and c 2 , the semantic relation between c 1 and c 2 is identical to the one between t 1 and t 2 because c 1 and c 2 contribute in the same way to the meanings of t 1 and t 2 . For instance, the relation between the MWTs croissance de la population 'population growth' and diminution de la population 'population decline' is the same as the one between the SWTs croissance 'growth' and diminution 'decline', that is antonymy. Our hypothesis is actually a bit stronger because we consider that the equivalence holds even when t 1 and t 2 do not have the same (syntactic) structure. More formally, let t 1 and t 2 be two MWTs such as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition method",
"sec_num": "3."
},
{
"text": "voc(t 1 ) = {u 1 , v 1 } and voc(t 2 ) = {u 2 , v 2 } where voc(x)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition method",
"sec_num": "3."
},
{
"text": "is the set of the content words of x. If u 1 and u 2 are SWTs, if v 1 = v 2 and if there is a semantic relation R between u 1 and u 2 , then R also holds between t 1 and t 2 . In other words, if M is a set of MWTs of a domain and S is a set of SWTs, the hypothesis can be stated as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition method",
"sec_num": "3."
},
{
"text": "\u2200t 1 \u2208 M, \u2200t 2 \u2208 M such as \u2203u 1 , v 1 , u 2 , v 2 / voc(t 1 ) = {u 1 , v 1 } \u2227 voc(t 2 ) = {u 2 , v 2 } \u2227u 1 \u2208 S \u2227 u 2 \u2208 S, [v 1 = v2 \u2227 \u2203R, R(u 1 , u 2 ) \u21d2 R(t 1 , t 2 )]",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition method",
"sec_num": "3."
},
{
"text": "4. Data and resources",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition method",
"sec_num": "3."
},
{
"text": "The corpus used for extracting MWT candidates is a specialized monolingual French corpus in the environment domain (ELRA-W0065) created in the framework of the PANACEA project 1 . The corpus contains 35453 documents (about 50 million words) with different levels of specialization. The corpus has been preprocessed: extraction of the text, normalization of the characters, lemmatization with TreeTagger (Schmid, 1994) .",
"cite_spans": [
{
"start": 403,
"end": 417,
"text": "(Schmid, 1994)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4.1."
},
{
"text": "The MWT candidates were extracted from the PANACEA corpus through TermSuit, a terminology extraction tool developed at LS2N 2 (Cram and Daille, 2016) . TermSuit only extracts noun phrases; the candidates are provided with their part of speech, specificity, and frequency. 1. Quasi-synonyms (QSYN): synonyms (diesel 'diesel' \u2194 gazole 'diesel'); quasi-synonyms (conserver 'preserve' \u2194 prot\u00e9ger 'protect'); close meanings (\u00e9lectricit\u00e9 'electricity' \u2194 \u00e9nergie 'energy'); variants (autopartage 'car sharing' \u2194 auto-partage 'car sharing').",
"cite_spans": [
{
"start": 126,
"end": 149,
"text": "(Cram and Daille, 2016)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "2. Hierarchical relations (HYP): hyponyms (autoroute 'highway' \u2192 route 'road'); hypernyms (combustible 'fuel' \u2192 p\u00e9trole 'oil'). Because HYP mixes hyponyms and hypernyms, the pairs it connects are not in order.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "3. Opposites (ANTI): antonyms (acc\u00e9l\u00e9rer 'accelerate' \u2194 ralentir 'slow down'); contrastives (flore 'flora' \u2194 faune 'fona').",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "4. Derivatives (DRV): terms with the same meaning but different parts of speech (sensibilit\u00e9 'sensitivity' \u2194 sensible 'sensible').",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "Because we are focusing on the projection of lexicalsemantic relations, we did not use the 259 DRV pairs and excluded them from RefCD. We also excluded the 225 pairs of verbs because TermSuite only extracts noun phrases.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "Since RefCD does not contain information between simple terms describing other relations, like co-hyponyms, our study on semantic relations between MWTs concentrates on QSYN, HYP, and ANTI. We extracted all the MWT candidates which contain two content words and formed all the MWTs pairs that share a common word and where the two other words are a pair of SWTs connected in RefCD. We did not impose any other restriction on PoS, the order of the constituents, nor the patterns of the MWT candidates. 18,382 pairs of MWT candidates have been created. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "TermSuite",
"sec_num": "4.2."
},
{
"text": "The raw projection yields symmetrical pairs of MWT candidates because some of the SWT pairs in RefCD are in random order. For instance, the projection produced the couple climat r\u00e9gional : climat local 'regional climate : local climate' and the couple climat local : climat r\u00e9gional. Therefore, we deleted the symmetries of hierarchical relationships. Table 4 shows the number of pairs that remained after the data filtering. ",
"cite_spans": [],
"ref_spans": [
{
"start": 352,
"end": 359,
"text": "Table 4",
"ref_id": "TABREF7"
}
],
"eq_spans": [],
"section": "Data filtering",
"sec_num": "5.2."
},
{
"text": "In order to assess the hypothesis that MWT meaning is compositional and that semantic relations between SWTs are preserved when they are projected on MWTs, we performed a manual validation on a subset of the MWT candidate pairs we have extracted. Since our study focuses on the preservation and the validity of the semantic relations, we do not want to include the quality of candidates in the validation (are they terms of the environmental field?). For instance, a candidate like lutte contre le changement 'fight against the change' is not a term because it is syntactically incomplete, and the actual term is lutte contre le changement climatique 'fight against climate change'. Additionally, a candidate like cadre r\u00e9gional 'regional framework' does not belong to the environment domain. Therefore, we choose to check the term status of the MWT candidates through three online terminological dictionaries, namely TERMIUM Plus 3 , Le Grand Dictionnaire 4 and IATE 5 (Interactive Terminology for Europe). We consider any candidate present in any of these resources is a term of the environmental field since it was extracted from a specialized corpus of this domain. Since many of the extracted terms are specific, such as conservation du papillon 'butterfly conservation', only a fraction of the pairs have both of their MWT candidates present in one of the resources. As shown in The meaning of a word strongly depends on the contexts where it is used. In this study, we show that the context also determines the meaning of MWTS and the relations that connect them. The annotation of the MWT pairs is based on the relation between the two SWTs they contain and five contexts (i.e., sentences) extracted from the corpus for each MWT. The validity of the projected relation is decided based on the meanings of the MWT occurrences in the extracted contexts. The relation is valid if it holds between the meanings of at least one occurrence of each of the MWTs. The context may help the judges understand the meaning of a MWT like zone de recharge 'recharge zone' which refers to a free aquifer where water collects. It can be used to disambiguate a term like air frais 'fresh air' which does not mean cool air but air from the outside (1). Contexts may also highlight the polysemy of MWTs like changement du climat 'climate change' which has two meanings: 'global warming' in (2a) and 'climate variability' in (2b).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Selection of a validation subset",
"sec_num": "5.3."
},
{
"text": "(1) la ventilation est \u00e0 double flux (l'air vici\u00e9 int\u00e9rieur r\u00e9chauffe l'air frais entrant)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Selection of a validation subset",
"sec_num": "5.3."
},
{
"text": "'the ventilation is double flow (the inside stale air heats the incoming fresh air)'",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Selection of a validation subset",
"sec_num": "5.3."
},
{
"text": "(2) a. il a \u00e9tabli que le changement du climat \u00e9tait \u00ab sans \u00e9quivoque \u00bb et que les \u00e9missions de gaz \u00e0 effet de serre provenant des activit\u00e9s humaines \u00e9taient responsables (avec 90% de certitude) de l'augmentation des temp\u00e9ratures depuis cent ans 'it established that the climate change was \"unequivocal\" and that greenhouse gas emissions from human activities were responsible (with 90% certainty) for the increase in temperatures over the past hundred years' b. \u00e0 quelle vitesse la r\u00e9duction des concentrations atmosph\u00e9riques de GES de courte dur\u00e9e entra\u00eenerait un changement du climat 'how quickly reducing short-lived atmospheric GHG concentrations would cause climate change'",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Selection of a validation subset",
"sec_num": "5.3."
},
{
"text": "The selected pairs have been annotated according to two criteria: the preservation of semantic relations and their validity in the environment domain. Both criteria are based on the expert knowledge of judges on semantic relations and the contexts in which the MWT candidates appear.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Criteria",
"sec_num": "6.2."
},
{
"text": "1. We consider that a relation is preserved when the relation that holds between two SWTs also holds between two MWT candidates generated from these two SWTs, regardless of its validity as an instance of its category. In other words, the relation is preserved when SWT 1 :SWT 2 ::MWT 1 :MWT 2 form an analogy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Criteria",
"sec_num": "6.2."
},
{
"text": "2. We consider that a relation between two MWT candidates is valid in the domain when it actually belongs to the category to which it is assigned.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Criteria",
"sec_num": "6.2."
},
{
"text": "We assessed the preservation of the relation and its validity separately because we have slightly changed the scope of the relation categories. We consider that co-hyponyms are not quasi-synonyms and cannot belong to QSYN. Furthermore, we consider the relationship between a pair of contrastive co-hyponym terms as an instance of ANTI.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Criteria",
"sec_num": "6.2."
},
{
"text": "The preservation of the relation only depends on the relations between the two SWTs and the two MWTs. If the relations are identical, the relation is considered as being preserved as in the case of temps froid : temps chaud 'cold weather : warm weather' (3) with respect to froid : chaud 'cold' : 'warm'.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Preservation",
"sec_num": "6.3."
},
{
"text": "(3) a. par temps froid, cette technique consiste \u00e0 ne pas laisser tourner son moteur au ralenti plus de 30 secondes 'by cold weather, this technique consists in not leaving the engine idling for more than 30 seconds' b. par temps chaud, le compromis entre confort et pratique est difficile \u00e0 trouver 'by warm weather, the compromise between comfort and practicality is difficult to find'",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Preservation",
"sec_num": "6.3."
},
{
"text": "On the other hand, diversit\u00e9 is a hypernym biodiversit\u00e9 in RefCD, but the contexts in (4) show that the relation between the MWTs gestion de la diversit\u00e9 'management of diversity' and gestion de la biodiversit\u00e9 'management of biodiversity' is different since they are used with the same meaning.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Preservation",
"sec_num": "6.3."
},
{
"text": "(4) a. les vari\u00e9t\u00e9s paysannes, issues de mill\u00e9naires de gestion de la diversit\u00e9 par les agriculteurs sont trop vivantes pour se plier aux crit\u00e8res d'inscription 'peasant varieties, coming from millennia of diversity management by farmers are too alive to comply with the criteria for registration' b. elle m\u00eame distincte de l'utilisation (par les agriculteurs) des semences, la gestion de la biodiversit\u00e9 cultiv\u00e9e r\u00e9unit dans un processus continu 'itself distinct from the use (by farmers) of seeds, the cultivated management of biodiversity unites in a continuous process'",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Preservation",
"sec_num": "6.3."
},
{
"text": "Relations that are not preserved are considered as invalid. However, not all preserved relations are valid in the domain. For instance, agriculture 'agriculture' is a hypernym of \u00e9levage 'lifestock farming' in RefCD, and the relation holding between these SWTs is preserved in the MWTs agriculture biologique 'organic agriculture' and \u00e9levage biologique 'organic lifestock farming'. However, a context like (5) shows that these MWTs are actually co-hyponyms because hypernyms cannot be coordinated in this way. The reason is that agriculture is polysemous and may also mean cultivation. In this context, agriculture and \u00e9levage are cohyponyms, and the inferred relation is not valid because it is not a relation of hypernymy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Domain validity",
"sec_num": "6.4."
},
{
"text": "(5) ... exp\u00e9rience avec une matrice agricole \"sans pesticides ni intrants chimiques\" (agriculture ou \u00e9levage biologique ou de prairies ... '... experience with an agricultural matrix \"without pesticides or chemical inputs\" (agriculture or organic farming or meadows ...'",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Domain validity",
"sec_num": "6.4."
},
{
"text": "Three judges have annotated the pairs of the validation subset. Table 7 shows that the inter-annotator agreement measured by Fleiss' kappa is substantial. The cases where the judges disagreed were then resolved. ANTI HYP QSYN 0.77 0.68 0.61 Table 7 : Fleiss' kappa",
"cite_spans": [],
"ref_spans": [
{
"start": 64,
"end": 71,
"text": "Table 7",
"ref_id": null
},
{
"start": 241,
"end": 248,
"text": "Table 7",
"ref_id": null
}
],
"eq_spans": [],
"section": "Analysis of the inferred relations",
"sec_num": "6.5."
},
{
"text": "The results (Table 8) show that most of MWTs have compositional meaning, which confirms the claim of (L'homme, 2004). They also show that the preservation and the validity of the projected relations vary with their category. Even if no restriction on the patterns was used for the generation of the MWT pairs, we observed that in all of the valid pairs, the MWTs have the same patterns and the SWTs that they contain appear in the same positions. 51 out of 231 pairs of MWTs are not preserved. They fall into three groups. (i) The MWTs do not have the same structure like eau de surface 'surface water' and surface de la terre 'Earth's surface'. eau 'water' and terre 'land' are linked by ANTI relation but the MWTs are not because eau and terre do not appear in the same position. (ii) The meaning of the SWTs is not preserved in the MWTs as in route maritime : autoroute maritime 'shipping route : marine highway'. route 'road' is a hypernym of autoroute 'highway', but route maritime and autoroute maritime are synonyms in the contexts extracted for these two MWTs. (iii) The change in meaning may also come from the content word shared by two MWTs as in air libre : eau libre 'outdoor : open water'. The 62 pairs where the relation has been considered invalid are mainly co-hyponyms formed by SWTs linked by a QSYN relation like trafic ferroviaire : trafic routier 'rail traffic : road traffic'.",
"cite_spans": [],
"ref_spans": [
{
"start": 12,
"end": 21,
"text": "(Table 8)",
"ref_id": "TABREF12"
}
],
"eq_spans": [],
"section": "Analysis of the inferred relations",
"sec_num": "6.5."
},
{
"text": "In this study, we have created a dataset of MWT pairs linked by semantic relations. These relations are projected from a reference list of SWTs connected by the same relations. The annotation of a subset of the data highlighted the importance of the contexts because they determine the real meaning of MWTs and subsequently, the semantic relation that holds between them. The following step in this research is to design a method to automate the annotation on the basis of the semantic relations between SWTs and contextual semantic model like BERT (Devlin et al., 2019) .",
"cite_spans": [
{
"start": 549,
"end": 570,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Works",
"sec_num": "7."
},
{
"text": "http://www.panacea-lr.eu/en/info-for-researchers/datasets/monolingual-corpora 2 https://www.ls2n.fr",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://www.btb.termiumplus.gc.ca/tpv2alpha/alphafra.html?lang=fra 4 http://www.granddictionnaire.com/ 5 https://iate.europa.eu/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Evaluation des modeles s\u00e9mantiques distributionnels: le cas de la d\u00e9rivation syntaxique",
"authors": [
{
"first": "G",
"middle": [],
"last": "Bernier-Colborne",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Drouin",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings the 23rd French Conference on Natural Language Processing (TALN)",
"volume": "",
"issue": "",
"pages": "125--138",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bernier-Colborne, G. and Drouin, P. (2016). Evaluation des modeles s\u00e9mantiques distributionnels: le cas de la d\u00e9rivation syntaxique. In Proceedings the 23rd French Conference on Natural Language Processing (TALN), pages 125-138.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Terminology extraction with term variant detection",
"authors": [
{
"first": "D",
"middle": [],
"last": "Cram",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Daille",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of ACL-2016 System Demonstrations",
"volume": "",
"issue": "",
"pages": "13--18",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Cram, D. and Daille, B. (2016). Terminology extraction with term variant detection. In Proceedings of ACL-2016 System Demonstrations, pages 13-18.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Semi-compositional method for synonym extraction of multi-word terms",
"authors": [
{
"first": "B",
"middle": [],
"last": "Daille",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Hazem",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14)",
"volume": "",
"issue": "",
"pages": "1202--1207",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Daille, B. and Hazem, A. (2014). Semi-compositional method for synonym extraction of multi-word terms. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14), pages 1202-1207, Reykjavik, Iceland, May. European Lan- guage Resources Association (ELRA).",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "BERT: Pre-training of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "J",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "M.-W",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "4171--4186",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2019). BERT: Pre-training of deep bidirectional trans- formers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics, pages 4171-4186, Minneapolis, Minnesota.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Detection of synonymy links between terms: experiment and results. Recent advances in computational terminology",
"authors": [
{
"first": "T",
"middle": [],
"last": "Hamon",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Nazarenko",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "2",
"issue": "",
"pages": "185--208",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hamon, T. and Nazarenko, A. (2001). Detection of syn- onymy links between terms: experiment and results. Re- cent advances in computational terminology, 2:185-208.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "A step towards the detection of semantic variants of terms in technical documents",
"authors": [
{
"first": "T",
"middle": [],
"last": "Hamon",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Nazarenko",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Gros",
"suffix": ""
}
],
"year": 1998,
"venue": "Proceedings of the 17th international conference on Computational linguistics",
"volume": "1",
"issue": "",
"pages": "498--504",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hamon, T., Nazarenko, A., and Gros, C. (1998). A step towards the detection of semantic variants of terms in technical documents. In Proceedings of the 17th interna- tional conference on Computational linguistics-Volume 1, pages 498-504. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Word embedding approach for synonym extraction of multi-word terms",
"authors": [
{
"first": "A",
"middle": [],
"last": "Hazem",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Daille",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the Eleventh International Conference on Language Resources and Evaluation",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hazem, A. and Daille, B. (2018). Word embedding ap- proach for synonym extraction of multi-word terms. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "La terminologie: principes et techniques",
"authors": [
{
"first": "M.-C",
"middle": [],
"last": "L'homme",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "L'homme, M.-C. (2004). La terminologie: principes et techniques. Pum.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Dicoenviro. dictionnaire fondamental de l'environnement",
"authors": [
{
"first": "M.-C",
"middle": [],
"last": "L'homme",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Lanneville",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "L'Homme, M.-C. and Lanneville, M. (2014). Dicoen- viro. dictionnaire fondamental de l'environnement. Consult\u00e9 \u00e0 l'adresse http://olst. ling. umontreal. ca/cgibin/dicoenviro/search. cgi.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Projecting corpus-based semantic links on a thesaurus",
"authors": [
{
"first": "E",
"middle": [],
"last": "Morin",
"suffix": ""
}
],
"year": 1999,
"venue": "Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "389--396",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Morin, E. (1999). Projecting corpus-based semantic links on a thesaurus. In Proceedings of the 37th Annual Meet- ing of the Association for Computational Linguistics, pages 389-396.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Learning antonyms with paraphrases and a morphology-aware neural network",
"authors": [
{
"first": "S",
"middle": [],
"last": "Rajana",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Callison-Burch",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Apidianaki",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Shwartz",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (* SEM 2017",
"volume": "",
"issue": "",
"pages": "12--21",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rajana, S., Callison-Burch, C., Apidianaki, M., and Shwartz, V. (2017). Learning antonyms with para- phrases and a morphology-aware neural network. In Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (* SEM 2017), pages 12-21.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Learning surface text patterns for a question answering system",
"authors": [
{
"first": "D",
"middle": [],
"last": "Ravichandran",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Hovy",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 40th annual meeting on association for computational linguistics",
"volume": "",
"issue": "",
"pages": "41--47",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ravichandran, D. and Hovy, E. (2002). Learning surface text patterns for a question answering system. In Pro- ceedings of the 40th annual meeting on association for computational linguistics, pages 41-47. Association for Computational Linguistics.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Probabilistic part-of-speech tagging using decision trees",
"authors": [
{
"first": "H",
"middle": [],
"last": "Schmid",
"suffix": ""
}
],
"year": 1994,
"venue": "Proceedings of the International Conference on New Methods in Language Processing",
"volume": "",
"issue": "",
"pages": "44--49",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Schmid, H. (1994). Probabilistic part-of-speech tagging using decision trees. In Proceedings of the International Conference on New Methods in Language Processing, pages 44-49.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Automatic discovery of similarwords",
"authors": [
{
"first": "P",
"middle": [],
"last": "Senellart",
"suffix": ""
},
{
"first": "V",
"middle": [
"D"
],
"last": "Blondel",
"suffix": ""
}
],
"year": 2008,
"venue": "Survey of Text Mining II",
"volume": "",
"issue": "",
"pages": "25--44",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Senellart, P. and Blondel, V. D. (2008). Automatic discov- ery of similarwords. In Survey of Text Mining II, pages 25-44. Springer.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Still a pain in the neck: Evaluating text representations on lexical composition",
"authors": [
{
"first": "V",
"middle": [],
"last": "Shwartz",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Dagan",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1902.10618"
]
},
"num": null,
"urls": [],
"raw_text": "Shwartz, V. and Dagan, I. (2019). Still a pain in the neck: Evaluating text representations on lexical composition. arXiv preprint arXiv:1902.10618.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Pattern-based synonym and antonym extraction. In Proceedings of the 48th annual southeast regional conference",
"authors": [
{
"first": "W",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Thomas",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Sheth",
"suffix": ""
},
{
"first": "Chan",
"middle": [],
"last": "",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "",
"suffix": ""
}
],
"year": 2010,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wang, W., Thomas, C., Sheth, A., and Chan, V. (2010). Pattern-based synonym and antonym extraction. In Pro- ceedings of the 48th annual southeast regional confer- ence, page 64. ACM.",
"links": null
}
},
"ref_entries": {
"TABREF0": {
"num": null,
"html": null,
"content": "<table><tr><td colspan=\"3\"># type pattern pilot</td><td>freq spec</td></tr><tr><td>3 T</td><td>N A</td><td>parc national</td><td>10198 4.17</td></tr><tr><td colspan=\"3\">3 V[s] N A A parc naturel national</td><td>59 1.94</td></tr><tr><td>4 T</td><td>A</td><td>communautaire</td><td>8864 4.11</td></tr><tr><td>13 T</td><td>T</td><td>biomasse</td><td>6239 3.96</td></tr><tr><td>17 T</td><td>N A</td><td>diversit\u00e9 biologique</td><td>5412 3.90</td></tr><tr><td>21 T</td><td>N A</td><td>milieu naturel</td><td>4328 3.80</td></tr><tr><td colspan=\"3\">21 V[s] N A A milieu naturel aquatique</td><td>23 1.54</td></tr></table>",
"text": "",
"type_str": "table"
},
"TABREF1": {
"num": null,
"html": null,
"content": "<table><tr><td>The semantic relations between MWT candidates are pre-</td></tr><tr><td>dicted from relations between SWTs. These semantically</td></tr><tr><td>linked SWTs are taken from a dataset made available by</td></tr><tr><td>Bernier-Colborne and Drouin (2016). This reference list</td></tr><tr><td>(RefCD) is extracted from DiCoEnviro (L'Homme and</td></tr><tr><td>Lanneville, 2014), a specialized dictionary of the environ-</td></tr><tr><td>ment field which describes the meaning of 1382 entry terms</td></tr><tr><td>of various sub-fields: energy, climate change, transporta-</td></tr><tr><td>tion, etc. RefCD is composed of 1314 term pairs, mainly</td></tr><tr><td>SWTs, connected by four relation categories:</td></tr></table>",
"text": "Excerpt of the TermSuite output 4.3. Reference list of linked terms",
"type_str": "table"
},
"TABREF2": {
"num": null,
"html": null,
"content": "<table><tr><td/><td>ANTI</td><td>HYP</td><td>QSYN</td><td>total</td></tr><tr><td>Pairs</td><td colspan=\"2\">116 191</td><td>523</td><td>830</td></tr><tr><td>Terms</td><td colspan=\"2\">107 122</td><td>415</td><td>429</td></tr></table>",
"text": "The distribution of the three relation categories is imbalanced, as shown in table 2.",
"type_str": "table"
},
"TABREF3": {
"num": null,
"html": null,
"content": "<table/>",
"text": "",
"type_str": "table"
},
"TABREF4": {
"num": null,
"html": null,
"content": "<table><tr><td>ANTI</td><td>HYP</td><td>QSYN</td><td>total</td></tr><tr><td colspan=\"4\">3414 3696 11,272 18,382</td></tr></table>",
"text": "presents their distribution over the three relation categories.",
"type_str": "table"
},
"TABREF5": {
"num": null,
"html": null,
"content": "<table/>",
"text": "MWTs yielded by the semantic projection",
"type_str": "table"
},
"TABREF7": {
"num": null,
"html": null,
"content": "<table/>",
"text": "",
"type_str": "table"
},
"TABREF8": {
"num": null,
"html": null,
"content": "<table><tr><td>ANTI</td><td>HYP</td><td>QSYN</td><td>total</td></tr><tr><td>80</td><td>51</td><td>100</td><td>231</td></tr></table>",
"text": "",
"type_str": "table"
},
"TABREF9": {
"num": null,
"html": null,
"content": "<table><tr><td>NA-NA</td><td>NA-NPN</td><td>NN-NN</td><td>NN-NPN</td><td>NPN-NPN</td></tr><tr><td>123</td><td>1</td><td>1</td><td>2</td><td>104</td></tr></table>",
"text": "Validation subsetphrases. In addition, most of the valid pairs are composed of two candidates having the same patterns, NA or NPN.",
"type_str": "table"
},
"TABREF10": {
"num": null,
"html": null,
"content": "<table><tr><td>: Distribution of pattern pairs of the validation sub-</td></tr><tr><td>set</td></tr><tr><td>6. Evaluation of semantic projection</td></tr><tr><td>6.1. Contexts</td></tr></table>",
"text": "",
"type_str": "table"
},
"TABREF12": {
"num": null,
"html": null,
"content": "<table/>",
"text": "Results of the validation",
"type_str": "table"
}
}
}
} |