File size: 76,434 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 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 | {
"paper_id": "2021",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T03:35:38.055494Z"
},
"title": "Adversarial Training for News Stance Detection: Leveraging Signals from a Multi-Genre Corpus",
"authors": [
{
"first": "Costanza",
"middle": [],
"last": "Conforti",
"suffix": "",
"affiliation": {
"laboratory": "Language Technology Lab",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Jakob",
"middle": [],
"last": "Berndt",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Mohammad",
"middle": [
"Taher"
],
"last": "Pilehvar",
"suffix": "",
"affiliation": {
"laboratory": "Language Technology Lab",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Marco",
"middle": [],
"last": "Basaldella",
"suffix": "",
"affiliation": {
"laboratory": "Language Technology Lab",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Chryssi",
"middle": [],
"last": "Giannitsarou",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Flavio",
"middle": [],
"last": "Toxvaerd",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
},
{
"first": "Nigel",
"middle": [],
"last": "Collier",
"suffix": "",
"affiliation": {
"laboratory": "Language Technology Lab",
"institution": "University of Cambridge",
"location": {}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Cross-target generalization constitutes an important issue for news Stance Detection (SD). In this short paper, we investigate adversarial cross-genre SD, where knowledge from annotated user-generated data is leveraged to improve news SD on targets unseen during training. We implement a BERT-based adversarial network and show experimental performance improvements over a set of strong baselines. Given the abundance of user-generated data, which are considerably less expensive to retrieve and annotate than news articles, this constitutes a promising research direction.",
"pdf_parse": {
"paper_id": "2021",
"_pdf_hash": "",
"abstract": [
{
"text": "Cross-target generalization constitutes an important issue for news Stance Detection (SD). In this short paper, we investigate adversarial cross-genre SD, where knowledge from annotated user-generated data is leveraged to improve news SD on targets unseen during training. We implement a BERT-based adversarial network and show experimental performance improvements over a set of strong baselines. Given the abundance of user-generated data, which are considerably less expensive to retrieve and annotate than news articles, this constitutes a promising research direction.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Stance Detection (SD) is an important NLP task (Mohammad et al., 2017) with widespread applications, ranging from rumor verification (Derczynski et al., 2017) and fact checking (Hanselowski et al., 2019) . Traditionally, research in SD focused on user-generated data, such as Twitter or Reddit (Gorrell et al., 2019) : this is mainly due to the abundance of such data, which are usually freely available online; moreover, user-generated data tend to be relatively short and compact, and thus more affordable to annotate and process. Starting from popular shared tasks such as Pomerleau and Rao (2017) , SD on complex and articulated input, such as news articles, has gained increasing popularity. Notably, effective news SD would constitute an invaluable tool to enhance the performance of human journalists in rumor and fake news debunking (Thorne and Vlachos, 2018) .",
"cite_spans": [
{
"start": 47,
"end": 70,
"text": "(Mohammad et al., 2017)",
"ref_id": "BIBREF14"
},
{
"start": 133,
"end": 158,
"text": "(Derczynski et al., 2017)",
"ref_id": "BIBREF5"
},
{
"start": 177,
"end": 203,
"text": "(Hanselowski et al., 2019)",
"ref_id": "BIBREF11"
},
{
"start": 294,
"end": 316,
"text": "(Gorrell et al., 2019)",
"ref_id": "BIBREF9"
},
{
"start": 576,
"end": 600,
"text": "Pomerleau and Rao (2017)",
"ref_id": "BIBREF15"
},
{
"start": 841,
"end": 867,
"text": "(Thorne and Vlachos, 2018)",
"ref_id": "BIBREF17"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In line with the general trend in NLP, deep learning-based models have long since established state-of-the-art results in news SD (Hanselowski et al., 2018) . Notably, training neural networks relies heavily on the availability of large labeled datasets, which are especially expensive to obtain for items such as news articles. As a consequence, following research on other text classification tasks such as sentiment analysis (Du et al., 2020) , research in SD investigated effective methods for cross-domain SD, where the scarcity of data for a specific dataset is supplemented with stance-annotated data from other domains. In this context, preliminary research in adversarial domain adaptation obtained promising results for both Twitter and news (Xu et al., 2019) SD.",
"cite_spans": [
{
"start": 130,
"end": 156,
"text": "(Hanselowski et al., 2018)",
"ref_id": "BIBREF10"
},
{
"start": 428,
"end": 445,
"text": "(Du et al., 2020)",
"ref_id": "BIBREF7"
},
{
"start": 752,
"end": 769,
"text": "(Xu et al., 2019)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this paper, we focus on the new task of crossgenre SD: we consider adversarial knowledge transfer from two datasets, WT-WT and STANDER, which collect samples in the same domain (i.e. the financial domain), but which belong to different genres (i.e. Twitter and news). We show experimentally that improvements in news SD performance can be achieved through cross-genre SD, which constitutes a promising direction for future research.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this work, we rely on two recently released datasets for news and Twitter SD: the STANDER corpus for the news genre (Conforti et al., 2020a) , and the WT-WT corpus for Twitter (Conforti et al., 2020b) . Both corpora collect samples discussing four mergers and acquisition (M&A) operations in the healthcare industry (Table 2) : an M&A operation, or merger, is the process in which a company (the buyer) attempts to acquire the ownership of another company (the target). A merger succeeds if ownership of the target is transferred, but can fail at any stage of discussions or can be blocked by authorities due to, e.g., antitrust concerns (Bruner and Perella, 2004) . Table 1 : Label distribution in the STANDER News SD corpus and in the WT-WT Twitter SD corpus.",
"cite_spans": [
{
"start": 119,
"end": 143,
"text": "(Conforti et al., 2020a)",
"ref_id": "BIBREF3"
},
{
"start": 179,
"end": 203,
"text": "(Conforti et al., 2020b)",
"ref_id": "BIBREF4"
},
{
"start": 641,
"end": 667,
"text": "(Bruner and Perella, 2004)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [
{
"start": 319,
"end": 328,
"text": "(Table 2)",
"ref_id": "TABREF2"
},
{
"start": 670,
"end": 677,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "An Aligned Multi-Genre Stance Detection Corpus",
"sec_num": "2"
},
{
"text": "Samples in both STANDER and WT-WT are manually stance-labeled by domain experts using a four-classes annotation schema distinguishing between support, refute, comment and unrelated, which expresses the sample's orientation about the outcome of the M&A (succeeded or rejected).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "An Aligned Multi-Genre Stance Detection Corpus",
"sec_num": "2"
},
{
"text": "As observed in Conforti et al. (2020a) , the two corpora present comparable signals, but display different characteristics which reflect the diverse genres they belong to. The Twitter samples are abundant and noisy, as indicated by the high percentage of unrelated and commenting samples ( Figure 1 and Table 1 ). On the other hand, STANDER collects considerably fewer samples, which are substantially longer and articulated; moreover, news articles in STANDER have been published in high-reputation outlets after careful editorial review, and thus contain a more formal and orthographically correct language with respect to user-generated tweets.",
"cite_spans": [
{
"start": 15,
"end": 38,
"text": "Conforti et al. (2020a)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [
{
"start": 290,
"end": 298,
"text": "Figure 1",
"ref_id": null
},
{
"start": 303,
"end": 310,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "An Aligned Multi-Genre Stance Detection Corpus",
"sec_num": "2"
},
{
"text": "Given the scarcity of news articles in STANDER, which are around one order of magnitude less abundant than tweets in WT-WT, and considering that both corpora collect the same targets in the same domain, cross-genre SD from WT-WT to STANDER seems to constitute an interesting research direction. However, due to the consistent genre differences, transferring knowledge from WT-WT to STANDER is non-trivial. To allow the model to capture the stance-specific features from the WT-WT samples which are useful to perform news SD, while ignoring the Twitter-specific features, we propose to treat the task adversarially.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Motivation",
"sec_num": "3.1"
},
{
"text": "We propose to consider two classification problems -SD and genre identification (GI) -with a shared BERT-based feature extractor, as shown in Figure 2 . To derive genre-invariant features, the GI component is trained adversarially. The model receives an input sample as:",
"cite_spans": [],
"ref_spans": [
{
"start": 142,
"end": 150,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "[CLS] Target [SEP] Text [SEP]",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": ", where Target is the SD target, expressed as the sentence \"A (a) will merge with B (b)\" (where upper-and lowercase a and b refers resp. to the buyer's and the target's company names and acronyms); for tweets, Text is the entire sample's text, while for news, we concatenate the article's title and its first four sentences into a single string. In this way, the target input is always the same over both genres, and it changes over targets only in the company names.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "Feature Extractor. As shared feature extractor, we adopt the pretrained BERT base uncased model (Devlin et al., 2019) .",
"cite_spans": [
{
"start": 96,
"end": 117,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "Stance Classifier. The stance label is predicted with a dense layer followed by a softmax operation operating on the hidden state h [CLS] of the special classification embedding [CLS]:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "y s = softmax(W s h [CLS] + b s )",
"eq_num": "(1)"
}
],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "The stance classifier is trained with categorical cross-entropy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "Genre Discriminator. The genre discriminator aims to predict gender labels of samples (Twitter or news). The feature extractor parameters are optimized to maximize the loss of the genre discriminator, thus encouraging BERT to generate genreinvariant features. In practice, the hidden state",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "h [CLS]",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "is first fed to a Gradient Reversal Layer (GRL, Ganin and Lempitsky (2014) ). During the forward propagation, the GRL acts as an identity transformation:",
"cite_spans": [
{
"start": 48,
"end": 74,
"text": "Ganin and Lempitsky (2014)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "GRL \u03bb (x) = x",
"eq_num": "(2)"
}
],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "but, during the backpropagation, it multiplies the gradient by a negative factor \u03bb:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "\u03b4GRL \u03bb (x)) \u03b4x = \u2212\u03bbI (3)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "The genre label y g is finally obtained with a dense layer followed by a sigmoid operation:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "y g = sigmoid(W g GRL(h [CLS] ) + b g )",
"eq_num": "(4)"
}
],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "The genre discriminator is trained with binary cross-entropy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "Joint Learning. The two components are jointly trained, resulting in the total loss:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "L total = L stance + L genre (5)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "The GI component is adversarial because it is trained to maximise the loss, while the SD component attempts to minimise it. In this way, the more the GI component is unable to correctly classify the samples, the more the system has learned to extract genre-invariant features.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "3.2"
},
{
"text": "Baselines. We report results with the three baseline models proposed in Conforti et al. (2020a) : a dummy random and majority vote baseline, abd BertEmb, an MLP leveraging sentence-BERT embeddings (Reimers and Gurevych, 2019) ; moreover, we also consider two further baselines:",
"cite_spans": [
{
"start": 72,
"end": 95,
"text": "Conforti et al. (2020a)",
"ref_id": "BIBREF3"
},
{
"start": 197,
"end": 225,
"text": "(Reimers and Gurevych, 2019)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "\u2022 BERT news : A vanilla BERT finetuned on news samples only; \u2022 BERT CoT rain : A vanilla BERT finetuned on Tweet and news samples, but without the adversarial component (Blum and Mitchell, 1998; Chen et al., 2011) .",
"cite_spans": [
{
"start": 169,
"end": 194,
"text": "(Blum and Mitchell, 1998;",
"ref_id": "BIBREF0"
},
{
"start": 195,
"end": 213,
"text": "Chen et al., 2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "Training Setting and Preprocessing. We train in a cross-target setting (train on three mergers, test on the fourth) with the Adam optimizer. For each configuration, we randomly select 20% of the training samples as heldout data. For experiments with adversarial cross-genre SD, we randomly select a number of Twitter samples equal to the news training samples (i.e. we double the size of the training set). This ratio was found to perform best in preliminary experiments (refer to the Appendix for further details). The test set contains news samples only. We lowercase both tweets and news samples.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "Hyperparameters. We set 128 as maximum sample length (including special tokens). We initialize our architecture with BERT large uncased 1 . BERT's weights are updated during training. We train using Adam (Kingma and Ba, 2015) on baatches of 23 samples, for a maximum of 70 epochs, with early stopping monitoring the SD loss on the development set.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "Evaluation. As in Conforti et al. (2020a,b) , and in line with other works on news SD (Hanselowski et al., 2018 (Hanselowski et al., , 2019 , we report on macro-averaged F 1 and consider both per-target operation scores, and average scores weighted by target operation size. For the adversarial cross-genre experiments, we compute accuracy for the binary GI task. For computing the evaluation metrics, we use the sklearn's implementations 2 .",
"cite_spans": [
{
"start": 18,
"end": 43,
"text": "Conforti et al. (2020a,b)",
"ref_id": null
},
{
"start": 86,
"end": 111,
"text": "(Hanselowski et al., 2018",
"ref_id": "BIBREF10"
},
{
"start": 112,
"end": 139,
"text": "(Hanselowski et al., , 2019",
"ref_id": "BIBREF11"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "Computing Infrastructure. We run experiments on an NVIDIA GeForce GTX 1080 GPU. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setting",
"sec_num": "4"
},
{
"text": "In this Section, we report on our cross-genre SD experiments. The discussion is organized around three research questions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "RQ1 What is the effect on news SD of including annotated data from a different genre?",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "As shown in Table 3 , BERT CoT rain performs better than all baselines, including the BERT news model, which was trained on in-genre data only. This seems to suggest that exposing the model to in-domain stance-annotated data, even if from a completely different genre, improves the generalizability over unseen targets.",
"cite_spans": [],
"ref_spans": [
{
"start": 12,
"end": 19,
"text": "Table 3",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "RQ2 Is adversarial training effective to improve cross-genre SD?",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "Adding an adversarial component to the BERT CoT rain model leads to gains in performance over all considered targets, with improvements ranging from +0.5 (AET HUM) and +4.0 (ANTM CI) in macro-averaged F 1 score. (Table 3) . Such performance gains are driven by improved performance over all stance labels (Table 4) , with refute samples benefiting most from the adversarial component. Such results suggest that a model which is punished for identifying the input's genre can still perform SD.",
"cite_spans": [],
"ref_spans": [
{
"start": 212,
"end": 221,
"text": "(Table 3)",
"ref_id": "TABREF4"
},
{
"start": 305,
"end": 314,
"text": "(Table 4)",
"ref_id": "TABREF6"
}
],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "RQ3 How does a decrease in GI through adversarial training correlate with SD performance? To understand to which extent genre-invariant representations are useful for SD, we experiment with different values of \u03bb, the GRL hyperparameter in Equation 3. As expected, GI performance lowers with increasing \u03bb (Table 4) , reaching 13.7 GI accuracy for the model with \u03bb = 1; this proves the GRL efficacy in forcing the model to learn genreindependent features. However, this also correlates with a steady decrease in SD performance, which holds true all target operations (Table 3) , with the only exception of the relatively small CI ESRX target, which also exhibit very strong label unbalancy (Table 1) .",
"cite_spans": [],
"ref_spans": [
{
"start": 304,
"end": 313,
"text": "(Table 4)",
"ref_id": "TABREF6"
},
{
"start": 565,
"end": 574,
"text": "(Table 3)",
"ref_id": "TABREF4"
},
{
"start": 688,
"end": 697,
"text": "(Table 1)",
"ref_id": null
}
],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "Moving to single-label classification, higher losses in performance are observed for comment and unrelated samples (resp. \u221213.5 and \u22128.3 in weighted accuracy), while support and refute label seem to be more robust to changes in the values of \u03bb. A possible explanation for this might be in the stylistic differences between the two corpora: unrelated and comment samples in the Twitter WT-WT corpus were often retrieved because of keywords homonymy 3 , and, as such, they tend to discuss completely different topics; on the contrary, such samples in the news STANDER corpus are actually covering the target companies. For this reason, completely genre-unaware knowledge transfer might not be optimal for those stance labels. This is in line with previous work by McHardy et al. (2019) on satire detection, and seems to indicate that, while learning partially genre-invariant features is beneficial for cross-target performance, features which are completely opaque with respect to the genre component are not ideal for SD.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Discussion",
"sec_num": "5"
},
{
"text": "In this paper, we discussed the new task of crossgenre SD: our experiments with a range of BERTbased architectures show that partially obfuscating the genre component through adversarial training leads to better generalization, especially considering low-frequency labels. Cross-genre SD thus constitutes a promising future research direction. Future work might include experiments using different underlying feature extractors, such as RoBERTa, or with adapters, to study the robustness of cross-genre SD over modeling choices. The integration of cross-genre and cross-domain adaptation, possibly in a multi-task setting as in Conforti et al. (2020a) , also offers interesting ideas for future investigation.",
"cite_spans": [
{
"start": 628,
"end": 651,
"text": "Conforti et al. (2020a)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "6"
},
{
"text": "In this Appendix, we report on the results of preliminary experiments which were run in order to find the best proportion between news and Twitter samples. We consider six settings, with increasing proportion of Twitter to news samples: 50%, 100%, 150%, 200%, 250% and a last setting in which all Twitter samples were included. For each setting, experiments were run in a cross-validation setting (training on data on three operations, and testing on the fourth). Table 5 : Stance Detection performance with the genreadversarial model (\u03bb = 0.2), by adding different proportion of Twitter data from the WT-WT corpus to the STANDER news samples. A % of 100 corresponds to the proportion used in the experiments reported in the paper. all corresponds to all the samples in the WT-WT corpus (33,668).",
"cite_spans": [],
"ref_spans": [
{
"start": 464,
"end": 471,
"text": "Table 5",
"ref_id": null
}
],
"eq_spans": [],
"section": "A Appendix",
"sec_num": null
},
{
"text": "Interestingly, we observe gains in overall performance w.r.t. the BERT news baseline, which is trained on news data only (Table 5) , with all adversarial models. This holds true even in the case of the model trained on the union of STANDER (2945) with all tweets from the WT-WT corpus (30711): the model's considerable performance gain with respect to the BERT news model testifies the ability of the adversarial model to learn partially genre-invariant features even when exposed to extremely unbalanced training data. Singlelabel results (Table 6) show that increasing the ratio of Twitter samples included in the training data tends to correlate with performance gains in recall, at the expense of losses in performance. The best news-to-tweets ratio lies between 100 and 250, with small differences between target operations and stance labels. Thus, our adversarial crossgenre models seem to be relatively robust over the exact amount of out-of-genre samples which are included during training. Table 6 : Per-label detailed performance when adding different percentage of tweets to the STANDER news samples. A % of 100 corresponds to the proportion used in the experiments reported in the paper. All results are obtained with the genre-adversarial model (\u03bb = 0.2). Considering macro-averaged Precision, Recall and F 1 measures, weighted according to the target operation's size.",
"cite_spans": [],
"ref_spans": [
{
"start": 121,
"end": 130,
"text": "(Table 5)",
"ref_id": null
},
{
"start": 540,
"end": 549,
"text": "(Table 6)",
"ref_id": null
},
{
"start": 999,
"end": 1006,
"text": "Table 6",
"ref_id": null
}
],
"eq_spans": [],
"section": "A Appendix",
"sec_num": null
},
{
"text": "https://tfhub.dev/tensorflow/bert en uncased L-12 H-768 A-12/3 2 https://scikit-learn.org/stable/modules/classes.html# module-sklearn.metrics",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "For example, 'cvs' not referring to the company 'CVS Health', but used as as plural of 'resume'.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "We thank the anonymous reviewers of this paper for their efforts and for the constructive comments and suggestions. We gratefully acknowledge funding from the Keynes Fund, University of Cambridge (grant no. JHOQ). CC is grateful to NERC DREAM CDT (grant no. 1945246) for partially funding this work. CG and FT are thankful to the Cambridge Endowment for Research in Finance (CERF).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Combining labeled and unlabeled data with co-training",
"authors": [
{
"first": "Avrim",
"middle": [],
"last": "Blum",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Tom",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mitchell",
"suffix": ""
}
],
"year": 1998,
"venue": "Proceedings of the Eleventh Annual Conference on Computational Learning Theory",
"volume": "",
"issue": "",
"pages": "92--100",
"other_ids": {
"DOI": [
"10.1145/279943.279962"
]
},
"num": null,
"urls": [],
"raw_text": "Avrim Blum and Tom M. Mitchell. 1998. Combin- ing labeled and unlabeled data with co-training. In Proceedings of the Eleventh Annual Conference on Computational Learning Theory, COLT 1998, Madison, Wisconsin, USA, July 24-26, 1998., pages 92-100.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Applied mergers and acquisitions",
"authors": [
{
"first": "F",
"middle": [],
"last": "Robert",
"suffix": ""
},
{
"first": "Joseph",
"middle": [
"R"
],
"last": "Bruner",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Perella",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "173",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Robert F Bruner and Joseph R Perella. 2004. Applied mergers and acquisitions, volume 173. John Wiley & Sons.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Co-training for domain adaptation",
"authors": [
{
"first": "Minmin",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Q",
"middle": [],
"last": "Kilian",
"suffix": ""
},
{
"first": "John",
"middle": [],
"last": "Weinberger",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Blitzer",
"suffix": ""
}
],
"year": 2011,
"venue": "Advances in Neural Information Processing Systems 24: 25th Annual Conference on Neural Information Processing Systems",
"volume": "",
"issue": "",
"pages": "2456--2464",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Minmin Chen, Kilian Q. Weinberger, and John Blitzer. 2011. Co-training for domain adaptation. In Ad- vances in Neural Information Processing Systems 24: 25th Annual Conference on Neural Information Processing Systems 2011. Proceedings of a meeting held 12-14 December 2011, Granada, Spain, pages 2456-2464.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "STANDER: an expertannotated dataset for news stance detection and evidence retrieval",
"authors": [
{
"first": "Costanza",
"middle": [],
"last": "Conforti",
"suffix": ""
},
{
"first": "Jakob",
"middle": [],
"last": "Berndt",
"suffix": ""
},
{
"first": "Mohammad",
"middle": [],
"last": "Taher Pilehvar",
"suffix": ""
},
{
"first": "Chryssi",
"middle": [],
"last": "Giannitsarou",
"suffix": ""
},
{
"first": "Flavio",
"middle": [],
"last": "Toxvaerd",
"suffix": ""
},
{
"first": "Nigel",
"middle": [],
"last": "Collier",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings, EMNLP 2020",
"volume": "",
"issue": "",
"pages": "4086--4101",
"other_ids": {
"DOI": [
"10.18653/v1/2020.findings-emnlp.365"
]
},
"num": null,
"urls": [],
"raw_text": "Costanza Conforti, Jakob Berndt, Mohammad Taher Pilehvar, Chryssi Giannitsarou, Flavio Toxvaerd, and Nigel Collier. 2020a. STANDER: an expert- annotated dataset for news stance detection and ev- idence retrieval. In Proceedings of the 2020 Con- ference on Empirical Methods in Natural Language Processing: Findings, EMNLP 2020, Online Event, 16-20 November 2020, pages 4086-4101. Associa- tion for Computational Linguistics.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Will-they-won't-they: A very large dataset for stance detection on twitter",
"authors": [
{
"first": "Costanza",
"middle": [],
"last": "Conforti",
"suffix": ""
},
{
"first": "Jakob",
"middle": [],
"last": "Berndt",
"suffix": ""
},
{
"first": "Mohammad",
"middle": [],
"last": "Taher Pilehvar",
"suffix": ""
},
{
"first": "Chryssi",
"middle": [],
"last": "Giannitsarou",
"suffix": ""
},
{
"first": "Flavio",
"middle": [],
"last": "Toxvaerd",
"suffix": ""
},
{
"first": "Nigel",
"middle": [],
"last": "Collier",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020",
"volume": "",
"issue": "",
"pages": "1715--1724",
"other_ids": {
"DOI": [
"10.18653/v1/2020.acl-main.157"
]
},
"num": null,
"urls": [],
"raw_text": "Costanza Conforti, Jakob Berndt, Mohammad Taher Pilehvar, Chryssi Giannitsarou, Flavio Toxvaerd, and Nigel Collier. 2020b. Will-they-won't-they: A very large dataset for stance detection on twitter. In Proceedings of the 58th Annual Meeting of the As- sociation for Computational Linguistics, ACL 2020, Online, July 5-10, 2020, pages 1715-1724. Associ- ation for Computational Linguistics.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Semeval-2017 task 8: Rumoureval: Determining rumour veracity and support for rumours",
"authors": [
{
"first": "Leon",
"middle": [],
"last": "Derczynski",
"suffix": ""
},
{
"first": "Kalina",
"middle": [],
"last": "Bontcheva",
"suffix": ""
},
{
"first": "Maria",
"middle": [],
"last": "Liakata",
"suffix": ""
},
{
"first": "Rob",
"middle": [],
"last": "Procter",
"suffix": ""
},
{
"first": "Geraldine",
"middle": [],
"last": "Wong Sak",
"suffix": ""
},
{
"first": "Arkaitz",
"middle": [],
"last": "Hoi",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Zubiaga",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the 11th International Workshop on Semantic Evaluation, SemEval@ACL 2017",
"volume": "",
"issue": "",
"pages": "69--76",
"other_ids": {
"DOI": [
"10.18653/v1/S17-2006"
]
},
"num": null,
"urls": [],
"raw_text": "Leon Derczynski, Kalina Bontcheva, Maria Liakata, Rob Procter, Geraldine Wong Sak Hoi, and Arkaitz Zubiaga. 2017. Semeval-2017 task 8: Rumoureval: Determining rumour veracity and support for ru- mours. In Proceedings of the 11th International Workshop on Semantic Evaluation, SemEval@ACL 2017, Vancouver, Canada, August 3-4, 2017, pages 69-76.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "BERT: pre-training of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "Jacob",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "Ming-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Kenton",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019",
"volume": "1",
"issue": "",
"pages": "4171--4186",
"other_ids": {
"DOI": [
"10.18653/v1/n19-1423"
]
},
"num": null,
"urls": [],
"raw_text": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: pre-training of deep bidirectional transformers for language under- standing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2-7, 2019, Volume 1 (Long and Short Pa- pers), pages 4171-4186. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Adversarial and domain-aware BERT for cross-domain sentiment analysis",
"authors": [
{
"first": "Chunning",
"middle": [],
"last": "Du",
"suffix": ""
},
{
"first": "Haifeng",
"middle": [],
"last": "Sun",
"suffix": ""
},
{
"first": "Jingyu",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Qi",
"middle": [],
"last": "Qi",
"suffix": ""
},
{
"first": "Jianxin",
"middle": [],
"last": "Liao",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "4019--4028",
"other_ids": {
"DOI": [
"10.18653/v1/2020.acl-main.370"
]
},
"num": null,
"urls": [],
"raw_text": "Chunning Du, Haifeng Sun, Jingyu Wang, Qi Qi, and Jianxin Liao. 2020. Adversarial and domain-aware BERT for cross-domain sentiment analysis. In Pro- ceedings of the 58th Annual Meeting of the Asso- ciation for Computational Linguistics, pages 4019- 4028, Online. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Unsupervised domain adaptation by backpropagation",
"authors": [
{
"first": "Yaroslav",
"middle": [],
"last": "Ganin",
"suffix": ""
},
{
"first": "Victor",
"middle": [],
"last": "Lempitsky",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1409.7495"
]
},
"num": null,
"urls": [],
"raw_text": "Yaroslav Ganin and Victor Lempitsky. 2014. Unsuper- vised domain adaptation by backpropagation. arXiv preprint arXiv:1409.7495.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "SemEval-2019 task 7: RumourEval, determining rumour veracity and support for rumours",
"authors": [
{
"first": "Genevieve",
"middle": [],
"last": "Gorrell",
"suffix": ""
},
{
"first": "Elena",
"middle": [],
"last": "Kochkina",
"suffix": ""
},
{
"first": "Maria",
"middle": [],
"last": "Liakata",
"suffix": ""
},
{
"first": "Ahmet",
"middle": [],
"last": "Aker",
"suffix": ""
},
{
"first": "Arkaitz",
"middle": [],
"last": "Zubiaga",
"suffix": ""
},
{
"first": "Kalina",
"middle": [],
"last": "Bontcheva",
"suffix": ""
},
{
"first": "Leon",
"middle": [],
"last": "Derczynski",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/S19-2147"
]
},
"num": null,
"urls": [],
"raw_text": "Genevieve Gorrell, Elena Kochkina, Maria Liakata, Ahmet Aker, Arkaitz Zubiaga, Kalina Bontcheva, and Leon Derczynski. 2019. SemEval-2019 task 7: RumourEval, determining rumour veracity and sup- port for rumours. In Proceedings of SemEval 2019.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "A retrospective analysis of the fake news challenge stance-detection task",
"authors": [
{
"first": "Andreas",
"middle": [],
"last": "Hanselowski",
"suffix": ""
},
{
"first": "P",
"middle": [
"V S"
],
"last": "Avinesh",
"suffix": ""
},
{
"first": "Benjamin",
"middle": [],
"last": "Schiller",
"suffix": ""
},
{
"first": "Felix",
"middle": [],
"last": "Caspelherr",
"suffix": ""
},
{
"first": "Debanjan",
"middle": [],
"last": "Chaudhuri",
"suffix": ""
},
{
"first": "Christian",
"middle": [
"M"
],
"last": "Meyer",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 27th International Conference on Computational Linguistics, COLING 2018",
"volume": "",
"issue": "",
"pages": "1859--1874",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andreas Hanselowski, Avinesh P. V. S., Benjamin Schiller, Felix Caspelherr, Debanjan Chaudhuri, Christian M. Meyer, and Iryna Gurevych. 2018. A retrospective analysis of the fake news challenge stance-detection task. In Proceedings of the 27th In- ternational Conference on Computational Linguis- tics, COLING 2018, Santa Fe, New Mexico, USA, August 20-26, 2018, pages 1859-1874. Association for Computational Linguistics.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "A richly annotated corpus for different tasks in automated factchecking",
"authors": [
{
"first": "Andreas",
"middle": [],
"last": "Hanselowski",
"suffix": ""
},
{
"first": "Christian",
"middle": [],
"last": "Stab",
"suffix": ""
},
{
"first": "Claudia",
"middle": [],
"last": "Schulz",
"suffix": ""
},
{
"first": "Zile",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL)",
"volume": "",
"issue": "",
"pages": "493--503",
"other_ids": {
"DOI": [
"10.18653/v1/K19-1046"
]
},
"num": null,
"urls": [],
"raw_text": "Andreas Hanselowski, Christian Stab, Claudia Schulz, Zile Li, and Iryna Gurevych. 2019. A richly anno- tated corpus for different tasks in automated fact- checking. In Proceedings of the 23rd Confer- ence on Computational Natural Language Learning (CoNLL), pages 493-503, Hong Kong, China. Asso- ciation for Computational Linguistics.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Adam: A method for stochastic optimization",
"authors": [
{
"first": "P",
"middle": [],
"last": "Diederik",
"suffix": ""
},
{
"first": "Jimmy",
"middle": [],
"last": "Kingma",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ba",
"suffix": ""
}
],
"year": 2015,
"venue": "3rd International Conference on Learning Representations",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In 3rd Inter- national Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Adversarial training for satire detection: Controlling for confounding variables",
"authors": [
{
"first": "Robert",
"middle": [],
"last": "Mchardy",
"suffix": ""
},
{
"first": "Heike",
"middle": [],
"last": "Adel",
"suffix": ""
},
{
"first": "Roman",
"middle": [],
"last": "Klinger",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019",
"volume": "1",
"issue": "",
"pages": "660--665",
"other_ids": {
"DOI": [
"10.18653/v1/n19-1069"
]
},
"num": null,
"urls": [],
"raw_text": "Robert McHardy, Heike Adel, and Roman Klinger. 2019. Adversarial training for satire detection: Con- trolling for confounding variables. In Proceed- ings of the 2019 Conference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies, NAACL- HLT 2019, Minneapolis, MN, USA, June 2-7, 2019, Volume 1 (Long and Short Papers), pages 660-665. Association for Computational Linguistics.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Stance and sentiment in tweets",
"authors": [
{
"first": "M",
"middle": [],
"last": "Saif",
"suffix": ""
},
{
"first": "Parinaz",
"middle": [],
"last": "Mohammad",
"suffix": ""
},
{
"first": "Svetlana",
"middle": [],
"last": "Sobhani",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Kiritchenko",
"suffix": ""
}
],
"year": 2017,
"venue": "ACM Trans. Internet Techn",
"volume": "17",
"issue": "3",
"pages": "",
"other_ids": {
"DOI": [
"10.1145/3003433"
]
},
"num": null,
"urls": [],
"raw_text": "Saif M. Mohammad, Parinaz Sobhani, and Svetlana Kiritchenko. 2017. Stance and sentiment in tweets. ACM Trans. Internet Techn., 17(3):26:1-26:23.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Fake news challenge",
"authors": [
{
"first": "Dean",
"middle": [],
"last": "Pomerleau",
"suffix": ""
},
{
"first": "Delip",
"middle": [],
"last": "Rao",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dean Pomerleau and Delip Rao. 2017. Fake news chal- lenge.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Sentencebert: Sentence embeddings using siamese bertnetworks",
"authors": [
{
"first": "Nils",
"middle": [],
"last": "Reimers",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nils Reimers and Iryna Gurevych. 2019. Sentence- bert: Sentence embeddings using siamese bert- networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Process- ing. Association for Computational Linguistics.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Automated fact checking: Task formulations, methods and future directions",
"authors": [
{
"first": "James",
"middle": [],
"last": "Thorne",
"suffix": ""
},
{
"first": "Andreas",
"middle": [],
"last": "Vlachos",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 27th International Conference on Computational Linguistics, COLING 2018",
"volume": "",
"issue": "",
"pages": "3346--3359",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "James Thorne and Andreas Vlachos. 2018. Automated fact checking: Task formulations, methods and fu- ture directions. In Proceedings of the 27th Inter- national Conference on Computational Linguistics, COLING 2018, Santa Fe, New Mexico, USA, Au- gust 20-26, 2018, pages 3346-3359. Association for Computational Linguistics.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Unseen target stance detection with adversarial domain generalization",
"authors": [
{
"first": "Zhen",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Qiansheng",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Chengguo",
"middle": [],
"last": "Lv",
"suffix": ""
},
{
"first": "Xue",
"middle": [],
"last": "Cao",
"suffix": ""
},
{
"first": "Guohong",
"middle": [],
"last": "Fu",
"suffix": ""
}
],
"year": 2020,
"venue": "2020 International Joint Conference on Neural Networks (IJCNN)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.1109/ijcnn48605.2020.9206635"
]
},
"num": null,
"urls": [],
"raw_text": "Zhen Wang, Qiansheng Wang, Chengguo Lv, Xue Cao, and Guohong Fu. 2020. Unseen target stance detec- tion with adversarial domain generalization. 2020 International Joint Conference on Neural Networks (IJCNN).",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Adversarial domain adaptation for stance detection",
"authors": [
{
"first": "Brian",
"middle": [],
"last": "Xu",
"suffix": ""
},
{
"first": "Mitra",
"middle": [],
"last": "Mohtarami",
"suffix": ""
},
{
"first": "James",
"middle": [],
"last": "Glass",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Brian Xu, Mitra Mohtarami, and James Glass. 2019. Adversarial domain adaptation for stance detection. CoRR, abs/1902.02401.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"num": null,
"uris": null,
"type_str": "figure",
"text": "Distribution and number of samples in the STANDER news SD and the WT-WT Twitter SD corpus(Figure takenfromConforti et al. (2020a)). Architecture of our model for news SD."
},
"TABREF2": {
"type_str": "table",
"text": "",
"html": null,
"content": "<table/>",
"num": null
},
"TABREF4": {
"type_str": "table",
"text": "Results on the STANDER target operations. Macro F 1 scores are obtained by testing on the target operation while training on the other three. Average scores are weighted by the target operation size.",
"html": null,
"content": "<table/>",
"num": null
},
"TABREF6": {
"type_str": "table",
"text": "",
"html": null,
"content": "<table><tr><td>: Per-target averaged accuracy for Gender Iden-</td></tr><tr><td>tification (GI) and per-target averaged F 1 score for</td></tr><tr><td>Stance Detection (SD), along with single-label per-</td></tr><tr><td>target averaged F 1 scores.</td></tr></table>",
"num": null
}
}
}
} |