File size: 77,493 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 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 | {
"paper_id": "2021",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T03:35:47.914862Z"
},
"title": "Zero-shot Cross-lingual Content Filtering: Offensive Language and Hate Speech Detection",
"authors": [
{
"first": "Andra\u017e",
"middle": [],
"last": "Pelicon",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jo\u017eef Stefan International Postgraduate School",
"location": {}
},
"email": "andraz.pelicon@ijs.si"
},
{
"first": "Ravi",
"middle": [],
"last": "Shekhar",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Mary University of London",
"location": {
"region": "Queen",
"country": "UK"
}
},
"email": "r.shekhar@qmul.ac.uk"
},
{
"first": "Matej",
"middle": [],
"last": "Martinc",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jo\u017eef Stefan International Postgraduate School",
"location": {}
},
"email": "matej.martinc@ijs.si"
},
{
"first": "",
"middle": [],
"last": "Bla\u017e\u0161krlj",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jo\u017eef Stefan International Postgraduate School",
"location": {}
},
"email": "blaz.skrlj@ijs.si"
},
{
"first": "Matthew",
"middle": [],
"last": "Purver",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jo\u017eef Stefan Institute",
"location": {
"settlement": "Ljubljana",
"country": "Slovenia"
}
},
"email": "m.purver@qmul.ac.uk"
},
{
"first": "Senja",
"middle": [],
"last": "Pollak",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Jo\u017eef Stefan Institute",
"location": {
"settlement": "Ljubljana",
"country": "Slovenia"
}
},
"email": "senja.pollak@ijs.si"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "We present a system for zero-shot crosslingual offensive language and hate speech classification. The system was trained on English datasets and tested on a task of detecting hate speech and offensive social media content in a number of languages without any additional training. Experiments show an impressive ability of both models to generalize from English to other languages. There is however an expected gap in performance between the tested cross-lingual models and the monolingual models. The best performing model (offensive content classifier) is available online as a REST API.",
"pdf_parse": {
"paper_id": "2021",
"_pdf_hash": "",
"abstract": [
{
"text": "We present a system for zero-shot crosslingual offensive language and hate speech classification. The system was trained on English datasets and tested on a task of detecting hate speech and offensive social media content in a number of languages without any additional training. Experiments show an impressive ability of both models to generalize from English to other languages. There is however an expected gap in performance between the tested cross-lingual models and the monolingual models. The best performing model (offensive content classifier) is available online as a REST API.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Recent years have seen a dramatic improvement in natural language processing, with machine learning systems outperforming human performance on a number of benchmark language understanding tasks (Wang et al., 2019) . This impressive achievement is somewhat tempered by the fact that a large majority of these systems work only for English, while other less-resourced languages are neglected due to a lack of training resources. On the other hand, another recent development is the introduction of systems capable of zero-shot cross-lingual transfer learning by leveraging multilingual embeddings (Artetxe and Schwenk, 2019) . These systems can be trained on a language with available resources and employed on a less-resourced language without any additional language specific training.",
"cite_spans": [
{
"start": 194,
"end": 213,
"text": "(Wang et al., 2019)",
"ref_id": "BIBREF22"
},
{
"start": 595,
"end": 622,
"text": "(Artetxe and Schwenk, 2019)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this study we present an offensive language classifier available through a REST API which leverages the cross-lingual capabilities of these systems. Due to the exponential growth of social media content, the amount of offensive language and hate speech has seen a steep increase and its identification and removal is no longer manageable by traditional manual inspection of the content (Schmidt and Wiegand, 2017) . As a consequence, there is a need for a general model that could be used in content filtering systems to automatically detect such discourse.",
"cite_spans": [
{
"start": 389,
"end": 416,
"text": "(Schmidt and Wiegand, 2017)",
"ref_id": "BIBREF17"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Since the majority of research in the area of offensive language and hate speech detection is currently done in monolingual settings, we performed a preliminary study to assess the feasibility of the proposed zero-shot cross-lingual transfer for this task. Two approaches are tested in this study. The first uses multilingual Bidirectional Encoder Representations from Transformers (BERT, Devlin et al., 2019) . The second uses Language-Agnostic SEntence Representations (LASER, Artetxe and Schwenk, 2019), a system built specifically for zero-shot cross-lingual transfer using multilingual sentence embeddings. Our best performing model is available online and can be used for detecting offensive content in less-resourced languages with no available training data.",
"cite_spans": [
{
"start": 389,
"end": 409,
"text": "Devlin et al., 2019)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The large majority of research on hate speech is monolingual, with English still the most popular language due to data availability (Wulczyn et al., 2017; Davidson et al., 2017) , and a number of English-only shared tasks organized on the topic of hate or offensive speech (e.g., OffenseEval, Zampieri et al., 2019b) . Lately, the focus has been shifting to other languages, with several shared tasks organized that cover other languages besides English, e.g. OffenseEval 2020 (Zampieri et al., 2020) , EVALITA 2018 (Bai et al., 2018) and Ger-mEval 2018 (Wiegand et al., 2018) .",
"cite_spans": [
{
"start": 132,
"end": 154,
"text": "(Wulczyn et al., 2017;",
"ref_id": "BIBREF24"
},
{
"start": 155,
"end": 177,
"text": "Davidson et al., 2017)",
"ref_id": "BIBREF6"
},
{
"start": 280,
"end": 316,
"text": "OffenseEval, Zampieri et al., 2019b)",
"ref_id": null
},
{
"start": 477,
"end": 500,
"text": "(Zampieri et al., 2020)",
"ref_id": "BIBREF27"
},
{
"start": 516,
"end": 534,
"text": "(Bai et al., 2018)",
"ref_id": "BIBREF1"
},
{
"start": 554,
"end": 576,
"text": "(Wiegand et al., 2018)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2"
},
{
"text": "For example, the EVALITA 2018 shared task (Bai et al., 2018) covered hate speech in Italian social media, the GermEval 2018 (Wiegand et al., 2018) shared tasks explored automatic identification of offensive German Tweets, and Semeval 2019 task 5 (Basile et al., 2019) covered detection of hate speech against immigrants and women in Spanish and English Twitter. Schmidt and Wiegand (2017) ; Poletto et al. (2020) ; Vidgen and Derczynski (2020) provide excellent surveys of recent hate speech related datasets. Ousidhoum et al. (2019) conduct multilingual hate speech studies by testing a number of traditional bag-of-words and neural models on a multilingual dataset containing English, French and Arabic tweets that were manually labeled with six class hostility labels (abusive, hateful, offensive, disrespectful, fearful, normal). They report that multilingual models outperform monolingual models on some of the tasks. Shekhar et al. (2020) study multilingual comment filtering for newspaper comments in Croatian and Estonian.",
"cite_spans": [
{
"start": 42,
"end": 60,
"text": "(Bai et al., 2018)",
"ref_id": "BIBREF1"
},
{
"start": 124,
"end": 146,
"text": "(Wiegand et al., 2018)",
"ref_id": "BIBREF23"
},
{
"start": 246,
"end": 267,
"text": "(Basile et al., 2019)",
"ref_id": "BIBREF3"
},
{
"start": 362,
"end": 388,
"text": "Schmidt and Wiegand (2017)",
"ref_id": "BIBREF17"
},
{
"start": 391,
"end": 412,
"text": "Poletto et al. (2020)",
"ref_id": "BIBREF15"
},
{
"start": 510,
"end": 533,
"text": "Ousidhoum et al. (2019)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2"
},
{
"text": "Another multilingual approach was proposed by Schneider et al. (2018) , who used multilingual MUSE embeddings (Lample et al., 2018) in order to extend the GermEval 2018 German train set with more English data. They report that no improvements in accuracy were achieved with this approach.",
"cite_spans": [
{
"start": 46,
"end": 69,
"text": "Schneider et al. (2018)",
"ref_id": "BIBREF18"
},
{
"start": 110,
"end": 131,
"text": "(Lample et al., 2018)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2"
},
{
"text": "Cross-lingual hate speech identification is even less researched than the multilingual task. The so-called bleaching approach (van der Goot et al., 2018) was used by Basile and Rubagotti (2018) to conduct cross-lingual experiments between Italian and English at EVALITA 2018 misogyny identification task. The only other study we are aware of is a very recent study by Pamungkas and Patti (2019) proposing an LSTM joint-learning model with multilingual MUSE embeddings. Google Translate is used for translation in order to create a bilingual train and test input data. Bassignana et al. (2018) report that the use of a multilingual lexicon of hate words, HurtLex, slightly improves the performance of misogyny identification systems. Closest to our work is that of Glava\u0161 et al. (2020) , who propose a dataset called XHATE-999 to evaluate abusive language detection in a multi-domain and multilingual setting.",
"cite_spans": [
{
"start": 166,
"end": 193,
"text": "Basile and Rubagotti (2018)",
"ref_id": "BIBREF2"
},
{
"start": 368,
"end": 394,
"text": "Pamungkas and Patti (2019)",
"ref_id": "BIBREF14"
},
{
"start": 568,
"end": 592,
"text": "Bassignana et al. (2018)",
"ref_id": "BIBREF4"
},
{
"start": 764,
"end": 784,
"text": "Glava\u0161 et al. (2020)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "2"
},
{
"text": "As an English (EN) training set for offensive language classification, we used the training subset of the OLID dataset (Zampieri et al., 2019a) . The trained models were evaluated on the test subset of the OLID dataset using their official gold labels and on the test subset of the GermEval 2018 dataset (Wiegand et al., 2018) , which also contains manually labeled tweets. Both datasets use hierarchical annotation schemes for annotating hate speech content. For our purposes, we employed only the annotations on the first level which classify tweets into two classes, offensive and not offensive.",
"cite_spans": [
{
"start": 119,
"end": 143,
"text": "(Zampieri et al., 2019a)",
"ref_id": "BIBREF25"
},
{
"start": 304,
"end": 326,
"text": "(Wiegand et al., 2018)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset Description",
"sec_num": "3"
},
{
"text": "We trained the hate speech classifiers on the English training set from the HatEval dataset (Basile et al., 2019) . For evaluation, we used the English and Spanish (ES) test sets from the HatEval competition, the German (DE) IGW hate speech dataset (Ross et al., 2016) , an Indonesian (ID) hate speech dataset (Ibrohim and Budi, 2019) and the Arabic (AR) hate speech dataset LHSAB (Mulki et al., 2019) . Each of the test datasets had binary labels that denoted the presence or absence of hate speech, except for the Arabic test set, which modeled hate speech as a three-class task, with labels denoting absence of hate speech, abusive language and hateful language. Since the authors themselves acknowledge there is a fine line between abusive and hateful language, we felt confident to join them into one class that denotes the presence of hate speech in a tweet. Tweets in the German IGW dataset included hate speech labels from two annotators and no common label, so we decided to evaluate only on those tweets where the two annotators agreed. The statistics of the datasets that were used in this study are reported in Table 1 .",
"cite_spans": [
{
"start": 92,
"end": 113,
"text": "(Basile et al., 2019)",
"ref_id": "BIBREF3"
},
{
"start": 249,
"end": 268,
"text": "(Ross et al., 2016)",
"ref_id": "BIBREF16"
},
{
"start": 310,
"end": 334,
"text": "(Ibrohim and Budi, 2019)",
"ref_id": "BIBREF10"
},
{
"start": 381,
"end": 401,
"text": "(Mulki et al., 2019)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [
{
"start": 1123,
"end": 1130,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Dataset Description",
"sec_num": "3"
},
{
"text": "Our models were trained and evaluated on two distinct albeit similar tasks, namely offensive language classification and hate speech detection, using two different approaches.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Classification models and methodology",
"sec_num": "4"
},
{
"text": "In the first approach, we tested the multilingual version of BERT to which we attached a classification layer with a softmax activation function. The model was fine-tuned on the chosen training datasets for 20 epochs. We limited the input sequence to 256 tokens and used a batch size of 32 and a learning rate of 2e-5. No additional hyperparameter tuning was performed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Classification models and methodology",
"sec_num": "4"
},
{
"text": "Our second approach was using the pre-trained LASER model and training a multilayer perceptron classifier with RELU activation function on top of that. To train the models we used the batch size of 32 and a learning rate of 0.001.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Classification models and methodology",
"sec_num": "4"
},
{
"text": "The results for both tasks together with the majority baselines and the results reported in the literature are presented in Table 2 . In the offensive language classification task, our best model (BERT) achieved an F1 score of 82.63 on the English test set, which is on par with the reported results achieved by monolingual classifiers (Zampieri et al., 2019b) . When evaluated on the German dataset, we observe a considerable drop in performance compared to the reported results (Wiegand et al., 2018) , however, it still achieves a solid F1 score of 70.67, which indicates its ability to generalize to languages it has not seen during training. In the hate speech classification task, the two models are comparable, with LASER outperforming BERT on the Arabic and Spanish datasets. Overall, the scores for the hate speech classification task proved to be considerably lower for both models as well as lower than the reported results in the monolingual experiments (Basile et al., 2019; Ibrohim and Budi, 2019) . Nevertheless, the results again indicate the ability of both models to generalize from English to other languages, as our models perform better than the majority baseline classifiers in terms of macro-averaged F1 score on all the datasets. It should be noted that the performance between our models and the reported performance on the Indonesian and Arabic datasets are not directly comparable as the original training and testing splits from the literature are not available. Therefore, our models were tested on different test splits.",
"cite_spans": [
{
"start": 336,
"end": 360,
"text": "(Zampieri et al., 2019b)",
"ref_id": "BIBREF26"
},
{
"start": 480,
"end": 502,
"text": "(Wiegand et al., 2018)",
"ref_id": "BIBREF23"
},
{
"start": 966,
"end": 987,
"text": "(Basile et al., 2019;",
"ref_id": "BIBREF3"
},
{
"start": 988,
"end": 1011,
"text": "Ibrohim and Budi, 2019)",
"ref_id": "BIBREF10"
}
],
"ref_spans": [
{
"start": 124,
"end": 131,
"text": "Table 2",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5"
},
{
"text": "The best performing cross-lingual model, multilingual BERT for offensive language classification, was implemented as a REST web service in the Flask framework. The design of the web service allows us to easily update the current model with a new version trained on additional data in the future. The web service can be reached programmatically through the endpoint at http:// classify.ijs.si/ml_hate_speech/ml_bert or through a demo browser-based interface at the URL http://classify.ijs.si/embeddia/ offensive_language_classifier. The interface is designed for mobile devices and supports most popular screen sizes. It consists of an input area where users can input their sentence and submit it for classification. The classification results as well as the confidence score of the classifier are then displayed under the input area.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Web API design",
"sec_num": "6"
},
{
"text": "In the course of this study, we tested the performance of two multilingual models, BERT and LASER, in zero-shot offensive language and hate speech detection. The results for the offensive language classification task show that even in the multilingual setting the BERT-based classifier achieves results comparable to the monolingual classifiers on English language data and solid performance on the German dataset. On the other hand, hate speech classification still proves to be a hard task for the multilingual classifiers as they achieve considerably lower scores on all languages compared to reported results. Nevertheless, both models show an impressive ability to generalize over languages they have not seen during fine-tuning. We implemented the best performing model, multilingual BERT for offensive language classification, as a REST web service. In the future, we plan to perform similar experiments with other multilingual language models, namely the XLM-R models (Conneau et al., 2019) , which show increased performance in standard benchmark tasks compared to multilingual BERT, and the recently released CroSloEngual-BERT (Ul\u010dar and Robnik-\u0160ikonja, 2020) .",
"cite_spans": [
{
"start": 976,
"end": 998,
"text": "(Conneau et al., 2019)",
"ref_id": "BIBREF5"
},
{
"start": 1137,
"end": 1169,
"text": "(Ul\u010dar and Robnik-\u0160ikonja, 2020)",
"ref_id": "BIBREF20"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and future work",
"sec_num": "7"
},
{
"text": "While all datasets used in this study contain social media posts labeled for hate speech or of- fensive language, there are still some differences in the way the data was labeled and collected, as each dataset was collected by a different research team. Therefore, some compromises had to be made in the course of this study to consolidate the datasets as best as possible. In order to better control for such variables, we would like to perform our experiment on the recently released XHate-999 dataset which contains instances in six diverse languages that were collected and annotated by the same research team using a unified annotation process. Given the fact we are working with relatively well-resourced languages, another future endeavour would be to also inspect the differences in cross-lingual model performance between zeroshot and few-shot testing scenarios. Finally, we plan on improving the performance of the model specifically on the task of hate speech classification, and update the existing web service.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and future work",
"sec_num": "7"
}
],
"back_matter": [
{
"text": "This research is supported by the European Union's Horizon 2020 research and innovation programme under grant agreement No 825153, project EM-BEDDIA (Cross-Lingual Embeddings for Less-Represented Languages in European News Media). The work of AP was funded also by the European Union's Rights, Equality and Citizenship Programme (2014-2020) project IMSyPP (Innovative Monitoring Systems and Prevention Policies of Online Hate Speech, grant no. 875263). The results of this publication reflect only the authors' views and the Commission is not responsible for any use that may be made of the information it contains. MP was also funded by the UK EPSRC under grant EP/S033564/1. We acknowledge also the funding by the Slovenian Research Agency (ARRS) core research programme Knowledge Technologies (P2-0103).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": "8"
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Massively multilingual sentence embeddings for zero-shot crosslingual transfer and beyond",
"authors": [
{
"first": "M",
"middle": [],
"last": "Artetxe",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Schwenk",
"suffix": ""
}
],
"year": 2019,
"venue": "Transactions of the ACL",
"volume": "7",
"issue": "",
"pages": "597--610",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00288"
]
},
"num": null,
"urls": [],
"raw_text": "M. Artetxe and H. Schwenk. 2019. Massively mul- tilingual sentence embeddings for zero-shot cross- lingual transfer and beyond. Transactions of the ACL, 7:597-610.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "RuG@EVALITA 2018: Hate speech detection in Italian social media",
"authors": [
{
"first": "X",
"middle": [],
"last": "Bai",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Merenda",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Zaghi",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Caselli",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Nissim",
"suffix": ""
}
],
"year": 2018,
"venue": "EVALITA Evaluation of NLP and Speech Tools for Italian",
"volume": "12",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "X. Bai, F. Merenda, C. Zaghi, T. Caselli, and M. Nis- sim. 2018. RuG@EVALITA 2018: Hate speech de- tection in Italian social media. EVALITA Evaluation of NLP and Speech Tools for Italian, 12:245.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "CrotoneMilano for AMI at Evalita2018. a performant, cross-lingual misogyny detection system",
"authors": [
{
"first": "A",
"middle": [],
"last": "Basile",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Rubagotti",
"suffix": ""
}
],
"year": 2018,
"venue": "EVALITA@ CLiC-it",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Basile and C. Rubagotti. 2018. CrotoneMilano for AMI at Evalita2018. a performant, cross-lingual misogyny detection system. In EVALITA@ CLiC-it.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "SemEval-2019 task 5: Multilingual detection of hate speech against immigrants and women in Twitter",
"authors": [
{
"first": "V",
"middle": [],
"last": "Basile",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Bosco",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Fersini",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Nozza",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Patti",
"suffix": ""
},
{
"first": "F",
"middle": [
"M"
],
"last": "Pardo",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Rosso",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Sanguinetti",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. SemEval",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/S19-2007"
]
},
"num": null,
"urls": [],
"raw_text": "V. Basile, C. Bosco, E. Fersini, D. Nozza, V. Patti, F. M. Rangel Pardo, P. Rosso, and M. Sanguinetti. 2019. SemEval-2019 task 5: Multilingual detection of hate speech against immigrants and women in Twitter. In Proc. SemEval.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Hurtlex: A multilingual lexicon of words to hurt",
"authors": [
{
"first": "E",
"middle": [],
"last": "Bassignana",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Basile",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Patti",
"suffix": ""
}
],
"year": 2018,
"venue": "5th Italian Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "E. Bassignana, V. Basile, and V. Patti. 2018. Hurtlex: A multilingual lexicon of words to hurt. In 5th Italian Conference on Computational Linguistics, CLiC-it 2018.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Unsupervised cross-lingual representation learning at scale",
"authors": [
{
"first": "A",
"middle": [],
"last": "Conneau",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Khandelwal",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Goyal",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Chaudhary",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Wenzek",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Guzm\u00e1n",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Grave",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Ott",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Stoyanov",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1911.02116"
]
},
"num": null,
"urls": [],
"raw_text": "A. Conneau, K. Khandelwal, N. Goyal, V. Chaud- hary, G. Wenzek, F. Guzm\u00e1n, E. Grave, M. Ott, L. Zettlemoyer, and V. Stoyanov. 2019. Unsuper- vised cross-lingual representation learning at scale. arXiv preprint arXiv:1911.02116.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Automated hate speech detection and the problem of offensive language",
"authors": [
{
"first": "T",
"middle": [],
"last": "Davidson",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Warmsley",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Macy",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Weber",
"suffix": ""
}
],
"year": 2017,
"venue": "Proc. ICWSM",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "T. Davidson, D. Warmsley, M. Macy, and I. Weber. 2017. Automated hate speech detection and the problem of offensive language. In Proc. ICWSM.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"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": "Proc. NAACL-HLT",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1423"
]
},
"num": null,
"urls": [],
"raw_text": "J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proc. NAACL-HLT.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "XHate-999: Analyzing and detecting abusive language across domains and languages",
"authors": [
{
"first": "G",
"middle": [],
"last": "Glava\u0161",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Karan",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Vuli\u0107",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 28th International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/2020.coling-main.559"
]
},
"num": null,
"urls": [],
"raw_text": "G. Glava\u0161, M. Karan, and I. Vuli\u0107. 2020. XHate-999: Analyzing and detecting abusive language across do- mains and languages. In Proceedings of the 28th International Conference on Computational Linguis- tics.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Bleaching text: Abstract features for cross-lingual gender prediction",
"authors": [
{
"first": "R",
"middle": [],
"last": "Van Der Goot",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Ljube\u0161i\u0107",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Matroos",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Nissim",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Plank",
"suffix": ""
}
],
"year": 2018,
"venue": "Proc. ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/P18-2061"
]
},
"num": null,
"urls": [],
"raw_text": "R. van der Goot, N. Ljube\u0161i\u0107, I. Matroos, M. Nissim, and B. Plank. 2018. Bleaching text: Abstract fea- tures for cross-lingual gender prediction. In Proc. ACL.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Multi-label hate speech and abusive language detection in Indonesian Twitter",
"authors": [
{
"first": "M",
"middle": [
"O"
],
"last": "Ibrohim",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Budi",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. 3rd Workshop on Abusive Language Online",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/W19-3506"
]
},
"num": null,
"urls": [],
"raw_text": "M. O. Ibrohim and I. Budi. 2019. Multi-label hate speech and abusive language detection in Indone- sian Twitter. In Proc. 3rd Workshop on Abusive Lan- guage Online.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Unsupervised machine translation using monolingual corpora only",
"authors": [
{
"first": "G",
"middle": [],
"last": "Lample",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Conneau",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Denoyer",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Ranzato",
"suffix": ""
}
],
"year": 2018,
"venue": "Proc. ICLR",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "G. Lample, A. Conneau, L. Denoyer, and M. Ran- zato. 2018. Unsupervised machine translation using monolingual corpora only. In Proc. ICLR.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "L-HSAB: A Levantine Twitter dataset for hate speech and abusive language",
"authors": [
{
"first": "H",
"middle": [],
"last": "Mulki",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Haddad",
"suffix": ""
},
{
"first": "C",
"middle": [
"Bechikh"
],
"last": "Ali",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Alshabani",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. 3rd Workshop on Abusive Language Online",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/W19-3512"
]
},
"num": null,
"urls": [],
"raw_text": "H. Mulki, H. Haddad, C. Bechikh Ali, and H. Alsha- bani. 2019. L-HSAB: A Levantine Twitter dataset for hate speech and abusive language. In Proc. 3rd Workshop on Abusive Language Online.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Multilingual and multi-aspect hate speech analysis",
"authors": [
{
"first": "N",
"middle": [],
"last": "Ousidhoum",
"suffix": ""
},
{
"first": "Z",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Song",
"suffix": ""
},
{
"first": "D.-Y",
"middle": [],
"last": "Yeung",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. EMNLP-IJCNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/D19-1474"
]
},
"num": null,
"urls": [],
"raw_text": "N. Ousidhoum, Z. Lin, H. Zhang, Y. Song, and D.- Y. Yeung. 2019. Multilingual and multi-aspect hate speech analysis. In Proc. EMNLP-IJCNLP.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Cross-domain and cross-lingual abusive language detection: A hybrid approach with deep learning and a multilingual lexicon",
"authors": [
{
"first": "E",
"middle": [
"W"
],
"last": "Pamungkas",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Patti",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. ACL Student Research Workshop",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/P19-2051"
]
},
"num": null,
"urls": [],
"raw_text": "E. W. Pamungkas and V. Patti. 2019. Cross-domain and cross-lingual abusive language detection: A hy- brid approach with deep learning and a multilingual lexicon. In Proc. ACL Student Research Workshop.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Resources and benchmark corpora for hate speech detection: a systematic review. Language Resources and Evaluation",
"authors": [
{
"first": "F",
"middle": [],
"last": "Poletto",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Basile",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Sanguinetti",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Bosco",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Patti",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "1--47",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "F. Poletto, V. Basile, M. Sanguinetti, C. Bosco, and V. Patti. 2020. Resources and benchmark corpora for hate speech detection: a systematic review. Lan- guage Resources and Evaluation, pages 1-47.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Measuring the Reliability of Hate Speech Annotations: The Case of the European Refugee Crisis",
"authors": [
{
"first": "B",
"middle": [],
"last": "Ross",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Rist",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Carbonell",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Cabrera",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Kurowsky",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Wojatzki",
"suffix": ""
}
],
"year": 2016,
"venue": "Proc. NLP4CMC III: 3rd Workshop on Natural Language Processing for Computer",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "B. Ross, M. Rist, G. Carbonell, B. Cabrera, N. Kurowsky, and M. Wojatzki. 2016. Measur- ing the Reliability of Hate Speech Annotations: The Case of the European Refugee Crisis. In Proc. NLP4CMC III: 3rd Workshop on Natural Lan- guage Processing for Computer-Mediated Commu- nication.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "A survey on hate speech detection using natural language processing",
"authors": [
{
"first": "A",
"middle": [],
"last": "Schmidt",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Wiegand",
"suffix": ""
}
],
"year": 2017,
"venue": "Proc. 5th International Workshop on Natural Language Processing for Social Media",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/W17-1101"
]
},
"num": null,
"urls": [],
"raw_text": "A. Schmidt and M. Wiegand. 2017. A survey on hate speech detection using natural language processing. In Proc. 5th International Workshop on Natural Lan- guage Processing for Social Media.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Towards the automatic classification of offensive language and related phenomena in German tweets",
"authors": [
{
"first": "J",
"middle": [
"M"
],
"last": "Schneider",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Roller",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Bourgonje",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Hegele",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Rehm",
"suffix": ""
}
],
"year": 2018,
"venue": "14th Conference on Natural Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. M. Schneider, R. Roller, P. Bourgonje, S. Hegele, and G. Rehm. 2018. Towards the automatic classifica- tion of offensive language and related phenomena in German tweets. In 14th Conference on Natural Lan- guage Processing KONVENS 2018.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Automating News Comment Moderation with Limited Resources: Benchmarking in Croatian and Estonian",
"authors": [
{
"first": "R",
"middle": [],
"last": "Shekhar",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Pranji\u0107",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Pollak",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Pelicon",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Purver",
"suffix": ""
}
],
"year": 2020,
"venue": "Journal for Language Technology and Computational Linguistics (JLCL)",
"volume": "",
"issue": "1",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R. Shekhar, M. Pranji\u0107, S. Pollak, A. Pelicon, and M. Purver. 2020. Automating News Comment Moderation with Limited Resources: Benchmarking in Croatian and Estonian. Journal for Language Technology and Computational Linguistics (JLCL), 34(1).",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "FinEst BERT and CroSloEngual BERT",
"authors": [
{
"first": "M",
"middle": [],
"last": "Ul\u010dar",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Robnik-\u0160ikonja",
"suffix": ""
}
],
"year": 2020,
"venue": "International Conference on Text, Speech, and Dialogue",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. Ul\u010dar and M. Robnik-\u0160ikonja. 2020. FinEst BERT and CroSloEngual BERT. In International Confer- ence on Text, Speech, and Dialogue.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Directions in abusive language training data, a systematic review: Garbage in, garbage out",
"authors": [
{
"first": "B",
"middle": [],
"last": "Vidgen",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Derczynski",
"suffix": ""
}
],
"year": 2020,
"venue": "Plos one",
"volume": "15",
"issue": "12",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "B. Vidgen and L. Derczynski. 2020. Directions in abusive language training data, a systematic review: Garbage in, garbage out. Plos one, 15(12):e0243300.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "GLUE: A multi-task benchmark and analysis platform for natural language understanding",
"authors": [
{
"first": "A",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Singh",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Michael",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Hill",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Levy",
"suffix": ""
},
{
"first": "S",
"middle": [
"R"
],
"last": "Bowman",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. ICLR",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Wang, A. Singh, J. Michael, F. Hill, O. Levy, and S. R. Bowman. 2019. GLUE: A multi-task bench- mark and analysis platform for natural language un- derstanding. In Proc. ICLR.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Overview of the GermEval 2018 shared task on the identification of offensive language",
"authors": [
{
"first": "M",
"middle": [],
"last": "Wiegand",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Siegel",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Ruppenhofer",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. Wiegand, M. Siegel, and J. Ruppenhofer. 2018. Overview of the GermEval 2018 shared task on the identification of offensive language.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Ex machina: Personal attacks seen at scale",
"authors": [
{
"first": "E",
"middle": [],
"last": "Wulczyn",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Thain",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Dixon",
"suffix": ""
}
],
"year": 2017,
"venue": "Proc. WWW",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.1145/3038912.3052591"
]
},
"num": null,
"urls": [],
"raw_text": "E. Wulczyn, N. Thain, and L. Dixon. 2017. Ex machina: Personal attacks seen at scale. In Proc. WWW.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Predicting the type and target of offensive posts in social media",
"authors": [
{
"first": "M",
"middle": [],
"last": "Zampieri",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Malmasi",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Farra",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Kumar",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. NAACL-HLT",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1144"
]
},
"num": null,
"urls": [],
"raw_text": "M. Zampieri, S. Malmasi, P. Nakov, S. Rosenthal, N. Farra, and R. Kumar. 2019a. Predicting the type and target of offensive posts in social media. In Proc. NAACL-HLT.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "SemEval-2019 task 6: Identifying and categorizing offensive language in social media (OffensEval)",
"authors": [
{
"first": "M",
"middle": [],
"last": "Zampieri",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Malmasi",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Farra",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Kumar",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. SemEval",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.18653/v1/S19-2010"
]
},
"num": null,
"urls": [],
"raw_text": "M. Zampieri, S. Malmasi, P. Nakov, S. Rosenthal, N. Farra, and R. Kumar. 2019b. SemEval-2019 task 6: Identifying and categorizing offensive language in social media (OffensEval). In Proc. SemEval.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Multilingual Offensive Language Identification in Social Media",
"authors": [
{
"first": "M",
"middle": [],
"last": "Zampieri",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Atanasova",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Karadzhov",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Mubarak",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Derczynski",
"suffix": ""
},
{
"first": "Z",
"middle": [],
"last": "Pitenis",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "\u00c7\u00f6ltekin",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "2020",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. Zampieri, P. Nakov, S. Rosenthal, P. Atanasova, G. Karadzhov, H. Mubarak, L. Derczynski, Z. Pite- nis, and c. \u00c7\u00f6ltekin. 2020. SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020). In Proceedings of SemEval.",
"links": null
}
},
"ref_entries": {
"TABREF1": {
"html": null,
"type_str": "table",
"content": "<table/>",
"num": null,
"text": "Results of the hate speech classification task (models trained on the English hatEval dataset) and offensive language classification task (models trained on the English OLID dataset) in comparison to the monolingual results as reported in the literature. The forward slash ('/') denotes results which are not reported in the literature. Figures marked with * denote results obtained on a different test split."
}
}
}
} |