File size: 110,951 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 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 | {
"paper_id": "2020",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T02:09:55.462220Z"
},
"title": "Annotating Topics, Stance, Argumentativeness and Claims in Dutch Social Media Comments: A Pilot Study",
"authors": [
{
"first": "Nina",
"middle": [],
"last": "Bauwelinck",
"suffix": "",
"affiliation": {
"laboratory": "LT3, Language and Translation Technology Team",
"institution": "Ghent University Groot",
"location": {
"addrLine": "Brittanni\u00eblaan 45",
"postCode": "9000",
"settlement": "Ghent",
"country": "Belgium"
}
},
"email": ""
},
{
"first": "Els",
"middle": [],
"last": "Lefever",
"suffix": "",
"affiliation": {
"laboratory": "LT3, Language and Translation Technology Team",
"institution": "Ghent University Groot",
"location": {
"addrLine": "Brittanni\u00eblaan 45",
"postCode": "9000",
"settlement": "Ghent",
"country": "Belgium"
}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "One of the major challenges currently facing the field of argumentation mining is the lack of consensus on how to analyse argumentative user-generated texts such as online comments. The theoretical motivations underlying the annotation guidelines used to generate labelled corpora rarely include motivation for the use of a particular theoretical basis. This pilot study reports on the annotation of a corpus of 100 Dutch user comments made in response to politically-themed news articles on Facebook. The annotation covers topic and aspect labelling, stance labelling, argumentativeness detection and claim identification. Our IAA study reports substantial agreement scores for argumentativeness detection (0.76 Fleiss' kappa) and moderate agreement for claim labelling (0.45 Fleiss' kappa). We provide a clear justification of the theories and definitions underlying the design of our guidelines. Our analysis of the annotations signal the importance of adjusting our guidelines to include allowances for missing context information and defining the concept of argumentativeness in connection with stance. Our annotated corpus and associated guidelines are made publicly available.",
"pdf_parse": {
"paper_id": "2020",
"_pdf_hash": "",
"abstract": [
{
"text": "One of the major challenges currently facing the field of argumentation mining is the lack of consensus on how to analyse argumentative user-generated texts such as online comments. The theoretical motivations underlying the annotation guidelines used to generate labelled corpora rarely include motivation for the use of a particular theoretical basis. This pilot study reports on the annotation of a corpus of 100 Dutch user comments made in response to politically-themed news articles on Facebook. The annotation covers topic and aspect labelling, stance labelling, argumentativeness detection and claim identification. Our IAA study reports substantial agreement scores for argumentativeness detection (0.76 Fleiss' kappa) and moderate agreement for claim labelling (0.45 Fleiss' kappa). We provide a clear justification of the theories and definitions underlying the design of our guidelines. Our analysis of the annotations signal the importance of adjusting our guidelines to include allowances for missing context information and defining the concept of argumentativeness in connection with stance. Our annotated corpus and associated guidelines are made publicly available.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "User-generated content (UGC) such as can be found in the comment sections of newspapers and social media sites is a valuable resource for the collection of argumentative texts written in natural language. According to Manosevitch and Walker (2009) , user comments offer a \"substantial amount of factual information, and [demonstrate] a public process of weighing alternatives via the expression of issue positions and supporting rationales\". The field of argumentation mining, which forms a part of Natural Language Processing (NLP) research, uses this type of data as a resource to train and test automatic detection systems for the purpose of extracting the various components making up the argumentation expressed by the users (Park and Cardie, 2014; Villalba and Saint-Dizier, 2012) . Training the systems requires annotating the data, for example labelling claims and reasons for those claims in the text. The various annotation tasks required for producing such data have proven to be very difficult for human annotators. Defining a good set of annotation guidelines is essential towards advancing the field of argumentation mining on UGC data such as social media comments. Currently, the myriad of theoretical perspectives on how to analyse argumentation as well as the unpredictable nature of UGC data have lead to a lack of current consensus on reliable guidelines for the various argumentation annotation tasks.",
"cite_spans": [
{
"start": 218,
"end": 247,
"text": "Manosevitch and Walker (2009)",
"ref_id": "BIBREF20"
},
{
"start": 730,
"end": 753,
"text": "(Park and Cardie, 2014;",
"ref_id": "BIBREF23"
},
{
"start": 754,
"end": 786,
"text": "Villalba and Saint-Dizier, 2012)",
"ref_id": "BIBREF34"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper presents a pilot annotation study for the identification of the topics, topic aspects and stance expressed by the comments, as well as the detection of argumentativeness and the main claim or conclusion presented. This study assesses the suitability of our current guidelines by measuring the Inter Annotator Agreement (IAA) for all tasks and analyses some specific cases which proved most challenging to our annotators. Our aim is to adjust the guidelines based on these results and analysis (Bauwelinck and Lefever, 2020) , which will then serve as the basis for an extensive annotation study on a more substantial corpus and including more annotation tasks required for a full analysis of the argumentation presented in the comments (a.o., this will include premise and argumentative relation annotation).",
"cite_spans": [
{
"start": 504,
"end": 534,
"text": "(Bauwelinck and Lefever, 2020)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In Section 2, we briefly discuss some of the relevant research. In Section 3, we give an overview of the theoretical frameworks which form the basis of our annotation scheme. In Section 4, we describe our pilot corpus and in Section 5 we give an overview of the annotation procedure, as well as more information on the rationale underlying our guidelines. In Section 6, we first present the results of the IAA study. In Section 7, we then present our analysis of the annotations. We end with Section 8 on concluding remarks as well as indications for future research.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In the field of argumentation mining, many different problems related to the analysis of argumentation in texts are being treated as different subtasks for automatic detection. Many of these tasks relate more generally to the processing of various aspects of texts within the broader field of NLP. For example, as a preliminary step towards the more argumentation-specific detection tasks, the tasks of topic and stance detection are often performed. Users, especially when arguing on controversial topics, tend to emphasize specific aspects of the topic, a concept called \"framing\" (Entman, 1993) . Therefore, both the more general topics and the more fine-grained topic aspects need to be identified. A major challenge still lies in determining how the different aspects relate to each other and to the major topic under discussion (Saint-Dizier, 2016) . This challenge relates to the issue of determining how fine-grained the targets of the users' stance needs to be. There is still no consensus on this issue, but the findings have confirmed that the more fine-grained the stance target, the more difficult it is to automatically classify the stance, and stance targets that are defined too broadly may not be specific enough to become associated with each respective side (pro/con) of the debate target (Wojatzki and Zesch, 2016) . Most authors therefore opt for a predetermined list of topics, in which topics either take the form of single words (more coarse-grained approach) or phrases (fine-grained). The latter approach was used by Saint-Dizier (2016), who defined controversial issues as evaluative statements (e.g., \"Climate action is necessary\") as this would aid in mining pro/con arguments for these specific controversies. As a preliminary step for the annotation of different argument components (Stab and Gurevych, 2014; Peldszus and Stede, 2015) , the text needs to be split into smaller segments. This step is often skipped in argumentation mining research, in favour of departing from pre-segmented text (Ajjour et al., 2017) . The segmentation of user-generated comments is not straightforward, as they contain many irregularities in the use of punctuation, capitalization and other orthographic markers of sentence boundaries. Determining the boundaries of argumentative segments is challenging but necessary for argumentation mining, as it forms the preliminary step both towards identifying the claim and premises as well as more fine-grained components of the argument. Textual indicators of argumentativeness such as discourse markers are often used as features for the automatic detection of argumentative segments (Eckle-Kohler et al., 2015) . An important caveat of using argumentative words or phrases for the task of argumentativeness detection is that they may also occur in non-argumentative text (van Eemeren and Houtlosser, 2006) . Argumentation mining research has produced some work on the linguistic cues which may be helpful for the identification of argumentativeness (Nguyen and Litman, 2015) , especially the detection of discourse markers has been explored to serve as signallers of argumentativeness to aid the automatic detection of arguments (Somasundaran et al., 2008; Tseronis, 2011; Eckle-Kohler et al., 2015) . The detection of segments representing the claim or conclusion of an argument in texts is an important prerequisite for applications involving fact checking (Vlachos and Riedel, 2014) . It is a very challenging task, especially when applied to an extremely varied resource, like online discourse . There is no current consensus on what exactly constitutes a claim, leading to many different annotation approaches (Daxenberger et al., 2017) .",
"cite_spans": [
{
"start": 583,
"end": 597,
"text": "(Entman, 1993)",
"ref_id": "BIBREF12"
},
{
"start": 834,
"end": 854,
"text": "(Saint-Dizier, 2016)",
"ref_id": "BIBREF27"
},
{
"start": 1308,
"end": 1334,
"text": "(Wojatzki and Zesch, 2016)",
"ref_id": "BIBREF37"
},
{
"start": 1814,
"end": 1839,
"text": "(Stab and Gurevych, 2014;",
"ref_id": "BIBREF30"
},
{
"start": 1840,
"end": 1865,
"text": "Peldszus and Stede, 2015)",
"ref_id": "BIBREF25"
},
{
"start": 2026,
"end": 2047,
"text": "(Ajjour et al., 2017)",
"ref_id": "BIBREF0"
},
{
"start": 2644,
"end": 2671,
"text": "(Eckle-Kohler et al., 2015)",
"ref_id": "BIBREF11"
},
{
"start": 2832,
"end": 2866,
"text": "(van Eemeren and Houtlosser, 2006)",
"ref_id": "BIBREF33"
},
{
"start": 3010,
"end": 3035,
"text": "(Nguyen and Litman, 2015)",
"ref_id": "BIBREF21"
},
{
"start": 3190,
"end": 3217,
"text": "(Somasundaran et al., 2008;",
"ref_id": "BIBREF29"
},
{
"start": 3218,
"end": 3233,
"text": "Tseronis, 2011;",
"ref_id": "BIBREF32"
},
{
"start": 3234,
"end": 3260,
"text": "Eckle-Kohler et al., 2015)",
"ref_id": "BIBREF11"
},
{
"start": 3420,
"end": 3446,
"text": "(Vlachos and Riedel, 2014)",
"ref_id": "BIBREF35"
},
{
"start": 3676,
"end": 3702,
"text": "(Daxenberger et al., 2017)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Research",
"sec_num": "2"
},
{
"text": "For the annotation of topics in comments, it is difficult to predetermine a list of possible topics. The comments in our corpus were made in response to newspaper articles shared via Facebook, so we can safely assume they will often refer to the newspaper content (Manosevitch and Walker, 2009) . In order to capture all the available topic information, the distinction between structuring and interactional topics is useful. The structuring topics are those found in the surrounding context; the interactional topics are those which form the topics of discussion and are found in the immediate context of interaction (Stromer-Galley and Martinson, 2009) . This distinction has been applied by Rowe (2015) , who used the two categories for the topic annotation of user comments to newspaper articles shared via Facebook and comments made on the same articles, via the official newspaper website. In their study, the structuring topic is that which is reported on in the article and the interactional topics are present in the individual comments. This approach seems feasible for our corpus, since the data type is so similar.",
"cite_spans": [
{
"start": 264,
"end": 294,
"text": "(Manosevitch and Walker, 2009)",
"ref_id": "BIBREF20"
},
{
"start": 618,
"end": 654,
"text": "(Stromer-Galley and Martinson, 2009)",
"ref_id": "BIBREF31"
},
{
"start": 694,
"end": 705,
"text": "Rowe (2015)",
"ref_id": "BIBREF26"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "For some interactional topics and aspects in the comment, the user expresses a stance reflecting his personal opinion towards the specific topic or aspect. Not all topics and aspects touched upon in the comment will be the target of the users' opinion, as some will only be mentioned in passing, or to set up the context for the argumentation. The typical stance annotation labels include pro, contra and none, where the last one represents a topic or aspect which is not used as a target for the opinion of the user (K\u00fc\u00e7\u00fck and Can, 2020) . Parallel to the distinction between the broader (discussion-wide) structuring topic label and the narrower interactional topic label (based on the comment text itself), the distinction between a debate stance and explicit stance is made by Wojatzki and Zesch (2016) . The debate stance is the stance towards the target of the whole debate. It is often implicit and inferrable from the explicit stance(s) which rely on textual evidence. This distinction has proven successful for stance annotation on noisy social media data and has even helped model implicit argumentation (Wojatzki and Zesch, 2016) .",
"cite_spans": [
{
"start": 517,
"end": 538,
"text": "(K\u00fc\u00e7\u00fck and Can, 2020)",
"ref_id": "BIBREF17"
},
{
"start": 781,
"end": 806,
"text": "Wojatzki and Zesch (2016)",
"ref_id": "BIBREF37"
},
{
"start": 1114,
"end": 1140,
"text": "(Wojatzki and Zesch, 2016)",
"ref_id": "BIBREF37"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "Given the difficulty of the automatic segmentation of argumentative texts (Al-Khatib et al., 2016) and our focus on the segment-level annotation of argumentativeness and claims, we also perform a preliminary manual segmentation step before handing the data to the annotators to avoid the error percolation which an added segmentation annotation task would inevitably introduce.",
"cite_spans": [
{
"start": 74,
"end": 98,
"text": "(Al-Khatib et al., 2016)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "While discourse markers are considered useful indicators of argumentativeness (Eckle-Kohler et al., 2015), Tseronis (2011) proposed the more specific concept of the argumentative marker which is a lexical item signalling the presence of an argumentative move (e.g., marking a standpoint). The argumentative marker may consist of a single word, phrase or sentence. A distinction is made between those markers which are syntactically part of the main constituents of the phrase and those that are independent on a semantic and syntactic level (Tseronis, 2011) . The concept of shell language developed by Madnani et al. (2012) is similar to the argumentative markers, in the sense that they may be used as signallers of argumentativeness and may consist of longer sequences of words than is the case for discourse markers. Shell language includes organizational phrases, such as ones marking the expression of an opinion (e.g., \"I think that\"), but also ones marking argumentative structures (such as \"after all\"). Du et al. (2014) have used shell language for the task of automatically separating topical contents from organizational language and have demonstrated the usefulness of this task, for instance to improve topic detection. They evaluated their fully unsupervised Shell Topic Model on argumentative UGC sourced from online forums. Ducrot (1982) contends that argumentativeness is present in an utterance even if it does not contain a linguistic expression it may explicitly be linked to. The first major challenge lies in defining what exactly constitutes argumentativeness. The distinction between the argumentative and informative components of utterances (Anscombre, 1995) provides one possible answer. The informative component corresponds to the propositional content of the sentence. The argumentative component signals the utterance's argumentative orientation: whether or not it has the potential of being used as a premise for a given conclusion. This perspective helps to circumvent the difficulty of reconstructing the intentions of the author of an argumentative text, as it emphasizes the text itself as the locus of the author's intention.",
"cite_spans": [
{
"start": 107,
"end": 122,
"text": "Tseronis (2011)",
"ref_id": "BIBREF32"
},
{
"start": 541,
"end": 557,
"text": "(Tseronis, 2011)",
"ref_id": "BIBREF32"
},
{
"start": 603,
"end": 624,
"text": "Madnani et al. (2012)",
"ref_id": "BIBREF19"
},
{
"start": 1013,
"end": 1029,
"text": "Du et al. (2014)",
"ref_id": "BIBREF9"
},
{
"start": 1341,
"end": 1354,
"text": "Ducrot (1982)",
"ref_id": "BIBREF10"
},
{
"start": 1668,
"end": 1685,
"text": "(Anscombre, 1995)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "Biran and Rambow (2011)'s concept of the claim consists of any utterance conveying subjective information and anticipating the question \"why are you telling me that?\". Daxenberger et al. (2017) have criticized Biran and Rambow's (2011) annotated dataset of online comments from blog threads for containing noisy sentences annotated with claims. However, we hypothesize this is not necessarily caused by the definition of the claim concept. Instead, it is a characteristic of this type of data. Our annotation guidelines employ a very similar definition of the claim concept as the one used in Biran and Rambow (2011) .",
"cite_spans": [
{
"start": 168,
"end": 193,
"text": "Daxenberger et al. (2017)",
"ref_id": "BIBREF8"
},
{
"start": 210,
"end": 235,
"text": "Biran and Rambow's (2011)",
"ref_id": "BIBREF5"
},
{
"start": 593,
"end": 616,
"text": "Biran and Rambow (2011)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "The current pilot study represents the first phase of our aim to define the annotation guidelines on a corpus of 100 Dutch user-generated comments sourced from the social media platform Facebook. All comments were made in response to an online newspaper article being shared via the official Facebook page of a Flemish newspaper. In this first phase, we focus on measuring agreement and finding edge cases for the annotation tasks of topic and stance labelling, segmentation of the text into argumentative units and claim identification. The second phase will consist of an agreement study on the same corpus sample of 100 comments and will focus on the tasks of premise and argumentative relation identification.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Theoretical Frameworks",
"sec_num": "3"
},
{
"text": "Facebook is by far the most popular platform for accessing news (Rowe, 2015) . However, given the difficulty in collecting this data automatically and assumptions about the lack of argumentation in shorter texts , this platform is rarely used to source data for argumentation mining purposes. Our corpus of 100 Dutch comments was sourced from the official Facebook page of Het Laatste Nieuws (HLN), a popular Flemish newspaper. As is the case for many news outlets (Rowe, 2015) , the Facebook page is used to share news articles with a wider audience, often linking to the official website of the newspaper. The corpus was collected manually and contains comments made on posts published in the second and third weeks of June 2020. We did not predetermine a list of topics to filter the Facebook posts. Instead, we chose to collect the first ten comments of each most recent post on the page which was topically related to a political or controversial topic. This included news stories related to policy decisions, party politics, but also related to topics like health care, poverty and racism. We filtered out duplicate comments and comments in other languages than Dutch, but did not filter out multiple comments made by the same user. Aside from the 100 comments, we also collected the 30 associated article texts and gathered 30 screenshots of the Facebook posts commented upon.",
"cite_spans": [
{
"start": 64,
"end": 76,
"text": "(Rowe, 2015)",
"ref_id": "BIBREF26"
},
{
"start": 465,
"end": 477,
"text": "(Rowe, 2015)",
"ref_id": "BIBREF26"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "For the segmentation of comments containing rhetorical questions, we decided to follow the perspective of Speech Act Theory, which contends that this special type of questions only has the surface structure of a question, but realizes the speech act of making a statement (Walton, 2007) . Thus, every rhetorical question which was followed by an explicit answer was considered as one segment together with that answer. Our approach resulted in a total number of 504 segments across the whole corpus. The highest number of segments for a single comment is 19; the lowest is 1. The comments in our corpus contained an average number of 5 segments.",
"cite_spans": [
{
"start": 272,
"end": 286,
"text": "(Walton, 2007)",
"ref_id": "BIBREF36"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "We divided our annotators into two groups of three. Each group was assigned different annotation tasks:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "1. The first group had to annotate the topics and topic aspects contained in the comment (interactional topics), as well as the stances expressed towards those topics and aspects. This task was performed on the level of the entire comment text. The annotators could select the topics and aspects from the list of structuring topics identified in the preliminary step for all the articles and Facebook posts. They were allowed to create additional topic and aspect labels if they could not find a suitable one in the list.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "2. The second group was tasked with labelling the predefined segments of each comment as argumentative unit (AU) or non-argumentative unit (NAU). Organizational elements which did not carry any argumentatively relevant information were to be marked as NAU. Then, the annotators had to indicate which segment best represented the claim of the argument. Only segments marked as AU were eligible to receive the claim label.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "We provide an example of an annotated comment from our corpus to help clarify the annotation tasks. We refer the reader to the guidelines for a more detailed description of the labels. For the comment \"And in the stores everyone touches all the fruit and vegetables with their hands better to monitor everything that is in the store, people even open packages and eat the product\", the topic-aspect combinations corona (measures, spread of disease), enforcement and shops were labelled by one of our annotators. Three stance labels were identified: pro towards the topic of enforcement and the aspect of measures and contra towards the aspect of spread of disease. The segmented comment was shown to the second group of annotators as follows: \"1[And] 2[in the stores everyone touches all the fruit and vegetables with their hands] 3[better to monitor everything that is in the store,] 4[people even open packages and eat the product]\". In a first round, the annotator determined for each segment whether it was argumentatively relevant or not, resulting in the following labels: argumentative (segments 2, 3 and 4); non-argumentative (segment 1). In a second round, the annotator was given the segments prelabelled for argumentativeness. The annotator then had to determine which of the segments labelled as argumentative best represented the central claim/conclusion of the user comment. In this example, segment 3 was labelled as the claim. All other segments marked argumentative in this comment are therefore seen to support or to be otherwise argumentatively relevant leading up to this claim.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Corpus",
"sec_num": "4"
},
{
"text": "For the first annotation tasks, viz. assigning topics, aspects and stance, the annotators could assign multiple labels to the same comment. As standard IAA scoring mechanisms (such as Cohen's pairwise kappa) assume the assignment of one category label per unit of annotation, these metrics are not suitable for measuring IAA for multiple labels per annotator. Therefore other metrics, such as Krippendorff's Alpha (Krippendorff, 1970; Krippendorff, 2004) , have to be used to calculate disagreement (or distance) between sets of assigned labels. Krippendorff's Alpha considers difference in annotation on all possible annotation units, irrespective of the number of labels and the type of annotation (categorical, numeric, ordinal). To calculate the distance between two sets of annotation labels, we followed the implementation of Passonneau (2006) , using MASI (Measuring Agreement on Set-valued Items). The resulting distance is 0 when sets are overlapping, and 1 when sets are disjoint.",
"cite_spans": [
{
"start": 414,
"end": 434,
"text": "(Krippendorff, 1970;",
"ref_id": "BIBREF15"
},
{
"start": 435,
"end": 454,
"text": "Krippendorff, 2004)",
"ref_id": "BIBREF16"
},
{
"start": 832,
"end": 849,
"text": "Passonneau (2006)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Inter-annotator Agreement Results",
"sec_num": "6"
},
{
"text": "The first annotation tasks appeared to be challenging, resulting in a total of 88 different topics, 245 aspects and 197 stance labels. As a lot of the disagreement for stance is caused by the choice of different Alpha/MASI distance #comments perfect agreement Ann1 -Ann2 Ann1 -Ann3 Ann2 -Ann3 Ann1 -Ann2 Ann1 -Ann3 Ann2 - aspects of the same topic, we also calculated the distance when only considering the sets of stance labels disregarding the specific topic or aspect they were attached to. Table 1 lists the distance between the sets of labels assigned per pair of annotators and the number of instances with total label agreement for topics, aspects, stance at the topic/aspect level and stance labels ignoring the aspect/topic. In the following step, annotators were charged with labeling each segment as (1) either an argumentative unit (AU) or non-argumentative unit (NAU) and (2) a claim or no claim. As annotators could only assign one label per task, we could apply more traditional IAA metrics for these tasks such as Cohen's pairwise kappa (Cohen, 1960) , which measures agreement between two raters, and Fleiss' kappa (Fleiss, 1971) , that can be used for measuring agreement between multiple raters. Note that the Fleiss kappa is a multi-rater generalization of Scott's pi statistic, not Cohen's kappa. Table 2 lists the agreement scores for these two tasks. For the interpretation of the kappa scores, we refer to Landis and Koch (1977) , who consider kappa scores ranging between 0.21 and 0.40 as fair agreement, between 0.41 and 0.60 as moderate agreement, between 0.61 and 0.80 as substantial agreement and between 0.81 and 0.99 as almost perfect agreement.",
"cite_spans": [
{
"start": 1053,
"end": 1066,
"text": "(Cohen, 1960)",
"ref_id": "BIBREF6"
},
{
"start": 1132,
"end": 1146,
"text": "(Fleiss, 1971)",
"ref_id": "BIBREF13"
},
{
"start": 1430,
"end": 1452,
"text": "Landis and Koch (1977)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [
{
"start": 494,
"end": 501,
"text": "Table 1",
"ref_id": null
},
{
"start": 1318,
"end": 1325,
"text": "Table 2",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Inter-annotator Agreement Results",
"sec_num": "6"
},
{
"text": "Fleiss' kappa Ann1 -Ann2 Ann1 -Ann3 Ann2 -Ann3 All annotators Argumentativeness /vs/ non-argumentativeness 0.74 0.82 0.71 0.76 Claim detection 0.45 0.48 0.41 0.45 We reached moderate agreement for the topic and aspect labelling tasks, considering the Krippendorff's distance is rather high between all pairs of annotators. This is partly due to errors percolating from the topic labelling step to the aspect labels. When we consider only the stance labels per comment, without taking into account the topic and aspects they are linked to, we see that on average, almost half the comments show perfect agreement. In addition, annotators sometimes assign similar (sub)topics or additional (sub)topics, resulting in fairly large distance scores: e.g. [travel, Corona, aviation] vs [Corona, aviation] results in a distance of 0.55, and [police, politician] vs [politician] results in a distance of 0.665. Perfect agreement on the stance labelling task was reached for only 8 comments, all three annotators agreeing on the \"none\" stance label for those comments. When considering partial agreement (defined here with the following condition satisfied: all three annotators share at least one stance label towards the same topic, e.g. \"(politics)contra\"), we noted partial agreement for 19 comments. In the following, we briefly discuss some concrete annotation examples for the annotation tasks which were performed on the comment level. When considering the distance metric between the stance labels only, we noted some recurring labelling errors.",
"cite_spans": [
{
"start": 748,
"end": 774,
"text": "[travel, Corona, aviation]",
"ref_id": null
},
{
"start": 778,
"end": 796,
"text": "[Corona, aviation]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "(1) Feeling ill = stay at home! Easy! There are people who can't stand wearing the mask due to breathing problems! Wearing a mask in this heat is asking for trouble for those people. Keep your distance, sanitize your hands and stay AT HOME when ill. EASY... 2",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "In some cases, the annotator identified the stance towards a specific topic, while the others identified only the stance towards the broader related topic. In Example 1, Annotator 1 identified a contra stance towards the mouth mask obligation topic, while Annotator 2 identified only a pro stance towards the more general topic of corona measures. Annotator 3 did identify both stances. One possible explanation for this confusion is that the annotators may have made a distinction between main topics and more peripheral topics in the comment. In future guidelines, it will be necessary to emphasize the need for identifying both the stance towards the broader topics and more specific ones.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "(2) Just have them all write a protest letter, you won't be receiving much mail.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "Comments containing irony such as Example 2 seem to complicate the stance annotation in some cases (perfect agreement for stance labelling amounting only to 8% of the total of 25 ironic comments and partial agreement reaching 28% (7/25)). Two of the annotators indicated doubts in cases such as these and were subsequently instructed to use the NONE label in case of doubt. This instruction will also be added to the future guidelines. Additionally, we will ask the annotators to mark the instances of irony and context-related understandability issues, a strategy also applied by Wojatzki and Zesch (2016) in their stance annotation study.",
"cite_spans": [
{
"start": 581,
"end": 606,
"text": "Wojatzki and Zesch (2016)",
"ref_id": "BIBREF37"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "(3) And the people who kept working from home every day and entertained the kids at the same time, also don't get anything from the government. The energy bill has nevertheless seriously increased in the past 3 months .. This decision was not well considered at all! Some specific topic labels frequently caused disagreement. In Example 3, disagreement is caused by the fact that one annotator has interpreted the user's stance as being in favour of the government handing out a financial compensation to those in need, resulting in the stance label (compensation)PRO. Another annotator has interpreted the stance as being against the lack of compensation for specific groups of the population, resulting in the label (compensation)CONTRA. The third annotator has circumvented the issue by adding the extra labels (government)CONTRA and (aid)PRO. In any case, the future guidelines will still include the option of defining extra labels wherever called for. Considering the difficulty of the annotation task, we reached substantial agreement scores for the task of argumentativeness detection (0.76 Fleiss' kappa) . For the argumentativeness labelling task, nonargumentative segments were defined as those having primarily an organizational function in the comment (for example, to introduce a particular component of the argumentation). Two annotators expressed their doubts as to how to annotate segments which in addition to the organizational function, also seemed to express the stance of the user, such as segments 2 and 6 in Example 4. To avoid mislabelling of such segments which are stance-bearing (and should thus be labelled as AU), we will add this specification to the guidelines. Segments occurring at the very start of the comment caused disagreement for argumentativeness labelling, especially in cases like Example 5 where the segment appeared to interact with the surrounding context (for instance, the title or the content of the newspaper article). Since the information in these segments often consists of an evaluative comment on the article content, they are considered stance-bearing and should be labelled as AU. We found a total of 45 comments in our corpus in which the first segment interacted with the context. In 77.8% of these cases (35/45 comments), perfect agreement was reached across all three annotators for the argumentativeness labelling task. When we compare this to the 52% of comments reaching perfect agreement for this task on the whole corpus, we see that the influence of the lack of context information available to our annotators was moderate for this task on our current corpus. We will ensure such segments are flagged in future annotations. This will help us to determine which types of comments may require context features for the automatic detection system.",
"cite_spans": [
{
"start": 1099,
"end": 1113,
"text": "Fleiss' kappa)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "Additionally, it will be useful to ensure our future guidelines are more explicit on the possible guises of argumentativeness (defined in our guidelines as \"all information relevant to the support or expression of the author's position\"). Therefore, we will explicitly state that this then includes background information (e.g., \"In 1400, a virus was considered an illness\") and (parts of) personal narratives (\"Just the other day I saw a whole family with no mouth masks on\"). In our corpus, we found a total of 44 comments contained at least one of these less obviously argumentative segments (corresponding to 19% or 98/504 total segments in our corpus). Since the annotators themselves raised the question of whether or not such segments were to be considered argumentative fairly quickly, we were able to achieve perfect agreement across all three annotators for 80% (79/98) of segments. However, perfect agreement was only reached for 43% (19/44) of all comments containing this type of segment, meaning that there were few comments in which they were all captured. Therefore, we will include more examples of what sequences of less obviously argumentative segments may look like (such as the first segment of Example (6) in the guidelines). In Example (6), while both segments are argumentative (since together they form evidence for the user's claim that the experts are distributing confusing information about the usefulness of mouth masks), we found disagreement on the argumentative relevance of the first segment.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "(6) 1[During the acute phase the experts said that wearing mouth masks wasn't useful .] 2[Now that the virus has been seriously reduced, we do have to wear them .] [...] Maar (English: but) (29 instances) and en (English: and) (23 instances) were the most frequently occurring shell language expressions and proved indicative of argumentativeness (But preceded an AU segment for 23/29 instances and And preceded an AU segment for 15/23 instances for all annotators). This is not surprising, since maar (English: but) is a connective often used to signal a contrasting reason in argumentation and en (English: and) can be used to signal an additional reason. An important type of shell language expression we found in our corpus (I shouldn't be saying this; There is not much more to say about this) corresponds to the so-called \"discourse markers of standpoint continuity\", identified by Craig and Sanusi (2000) as a common characteristic of group discussions on controversial issues. They are commonly used to specify argumentative standpoints as well as to avoid disagreement while saving face (Craig and Sanusi, 2000) . However, a clear distinction should be introduced in the guidelines between such markers and segments like I do understand his point or I get that, which are indicative of the stance of the author and should be annotated as AU (all three annotators currently annotated these segments as NAU). The presence of the verbs of understanding as well as the first person singular may help distinguish such segments.",
"cite_spans": [
{
"start": 164,
"end": 169,
"text": "[...]",
"ref_id": null
},
{
"start": 888,
"end": 911,
"text": "Craig and Sanusi (2000)",
"ref_id": "BIBREF7"
},
{
"start": 1096,
"end": 1120,
"text": "(Craig and Sanusi, 2000)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "We reached moderate agreement for the claim labelling task (0.45 Fleiss' kappa). For 33 comments full agreement was reached on the claim annotation task. Many of the claims identified are elliptical, e.g.: Belgium doomed, Mayor not capable. This is typical of the nature of our data. From Daxenberger et al.'s (2017) analysis of claim segments in various argumentation mining corpora, the presence of policy claims (Schiappa and Nordin, 2014) emerged as a common characteristic in multiple corpora (e.g., of the Wikipedia Talk Page Corpus (Biran and Rambow, 2011) and the Microtext corpus of Peldszus and Stede (2015) ). In our corpus, such policy claims, which are often characterized by the presence of the modal verb \"should\" (Daxenberger et al., 2017) , were present in 32 comments, for example: Everyone should just decide for themselves what they find important, taking care of yourself or blending in with the crowd. Since the policy claim expresses a wish for things to be done differently, it may be expressed in the form of an advice (We can stop driveling about the mouth masks now), or as a strong imperative (if your heart isn't in it, don't do it), the latter of which seems to be particularly indicative of claim segments. Perfect agreement for the claim labeling task was reached on 44% of comments (14/32) (corresponding to 22 segments containing at least one segment with \"should\" or an imperative form.",
"cite_spans": [
{
"start": 289,
"end": 316,
"text": "Daxenberger et al.'s (2017)",
"ref_id": "BIBREF8"
},
{
"start": 415,
"end": 442,
"text": "(Schiappa and Nordin, 2014)",
"ref_id": "BIBREF28"
},
{
"start": 539,
"end": 563,
"text": "(Biran and Rambow, 2011)",
"ref_id": "BIBREF5"
},
{
"start": 592,
"end": 617,
"text": "Peldszus and Stede (2015)",
"ref_id": "BIBREF25"
},
{
"start": 729,
"end": 755,
"text": "(Daxenberger et al., 2017)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "Aside from the occurrence of policy claims, when comparing the claims annotated in our corpus to those found in the Web Discourse corpus , our claims are more often anaphoric in the sense that they express the stance of the author, but without specific lexical reference to the given topic (\"I personally find it really sad it's not obligatory,\"). Many claims contain expressions signalling beliefs (such as \"in my opinion\", \"personally\", \"I find\"), which according to Daxenberger et al.'s (2017) analysis, is characteristic of the claims found in persuasive student essays (such as the corpus of Stab and Gurevych (2017) ). In general, the claim of an argument is more likely to carry stance-taking words toward the topic. This aspect has been identifed as a useful feature for the automatic detection of claims (Ajjour et al., 2019) .",
"cite_spans": [
{
"start": 469,
"end": 496,
"text": "Daxenberger et al.'s (2017)",
"ref_id": "BIBREF8"
},
{
"start": 597,
"end": 621,
"text": "Stab and Gurevych (2017)",
"ref_id": "BIBREF8"
},
{
"start": 813,
"end": 834,
"text": "(Ajjour et al., 2019)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "The fact that we can find many correspondences with existing corpora of different genres and domains, strengthens our assumption that general markers of claim presence may still be found across genres. Adding these markers in the guidelines for the annotation can help create a more unified approach towards the annotation of claims, which appeared to be absent in the field (Daxenberger et al., 2017) . We will investigate the presence and usefulness of such general markers of argumentativity in our more extended corpus, including more domains and platforms to source UGC data from.",
"cite_spans": [
{
"start": 375,
"end": 401,
"text": "(Daxenberger et al., 2017)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Cohen's kappa",
"sec_num": null
},
{
"text": "The insights we gained from this pilot annotation study will be used to improve our annotation guidelines. In this study, we pre-segmented all the comments in a preliminary step. This was done to avoid too much error percolation from the annotation results of the segmentation into the claim annotation task, which would make calculating the agreement on claim identification a difficult matter. However, we realize this pre-segmentation may introduce more bias when considering the annotation of other argumentative components like premises and relations between segments. Since the automatic segmentation of texts into argumentative and non-argumentative segments is still very challenging and advances in the field are still being developed, we will use Al-Khatib et al.'s (2016) rule-based algorithm for the automatic pre-segmentation of the corpus as a step prior to the manual annotation. In this way, the annotators will be asked to correct the automatic segmentation by merging incorrectly split segments.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "8"
},
{
"text": "We believe the low agreement results for the first set of annotation tasks (topic, aspect and stance detection) may be improved by reducing the number of possible structuring topic labels for the annotators to choose from. This will require pruning the list of structuring topics and aspects we identified to include only the most frequently occurring ones. Since we are aware of the risk of bias entering our annotation process in providing the structuring topic and aspect labels for the annotators to choose from when deciding on the interactional labels, we will include an evaluation step (removing duplicate labels and becoming aware of ambiguities in certain labels) in our revised annotation guidelines to be performed by a separate annotator. From our analysis of the results for the argumentativeness annotation task, we conclude that our guidelines need more incorporation of stance expressions as an important indicator of argumentativeness.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "8"
},
{
"text": "Since most of our annotators indicated that they had trouble annotating comments due to missing context, we want to explore the impact of context on the various annotation tasks we have performed in this study. First of all, our new guidelines will have to supply more examples of information that is considered argumentatively relevant, e.g., background or context setting information. In particular, we want to investigate whether there are \"triggering devices\" which are used to evoke context (Nyan, 2017) in the comment. Since the function of context is often defined as narrowing down the range of possible understandings of a text or utterance (Nyan, 2017) , we are interested in studying how supplying the annotators with various degrees of context information will for instance impact their understanding of the argumentativeness and the claim of a comment.",
"cite_spans": [
{
"start": 496,
"end": 508,
"text": "(Nyan, 2017)",
"ref_id": "BIBREF22"
},
{
"start": 650,
"end": 662,
"text": "(Nyan, 2017)",
"ref_id": "BIBREF22"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "8"
},
{
"text": "AnnotationSix annotators participated in this study. All of them were Dutch native speakers and linguists who received the same annotation guidelines(Bauwelinck and Lefever, 2020) to follow 1 . They each annotated a total of 100 Dutch user comments, all 100 were the same set for each annotator and were used to measure inter-annotator agreement.We performed two preliminary steps to prepare the corpus for the annotators. First, to prepare for the topic annotations we annotated the structuring topics and aspects contained in the Facebook posts and the news article texts. For the topic annotations of the Facebook posts, we labelled the topic information as present on screenshots of the post in question, thus including information like the title of the article, the accompanying image and description text in our decision making. For the topic annotations of the article texts, we limited ourselves to the title and the first three paragraphs of each article. The second preparatory step was to manually segment all the comments in order to prepare for the argumentativeness and claim annotation tasks, which were to be performed on a segment-level. We segmented the comments based on the shell language expressions we found(Madnani et al., 2012;Du et al., 2014), but only if they were also set apart from other segments in a syntactical way(Tseronis, 2011). If the expression occurred syntactically embedded in the phrase, we considered it a part of the larger segment. This approach allowed us to distinguish organizational segments from content segments. We did not consider markers of opinion (such as \"I don't think that\") as part of shell language, since we only wanted to focus on separating organizational phrases marking argumentative structures from content segments, leaving markers of subjectivity to remain a part of the content segments. Thus, our segmentation approach is more coarse-grained. Other researchers likeNguyen and Litman (2015) have used shell-like concepts for the task of detecting argument components in persuasive essays, also allowing shell language (inNguyen and Litman's (2015) terms \"argument words\") to occur in the argument content (inNguyen and Litman's (2015) terms \"domain words\").1 Both the corpus (consisting of comments, article texts and Facebook screenshots) and the associated annotation guidelines can be found at https://www.lt3.ugent.be/resources/platos pilot-study/.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "All examples from the corpus given here have been translated from the original Dutch.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Unit segmentation of argumentative texts",
"authors": [
{
"first": "Y",
"middle": [],
"last": "Ajjour",
"suffix": ""
},
{
"first": "W",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Kiesel",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Wachsmuth",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Stein",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the 4th Workshop on Argument Mining",
"volume": "",
"issue": "",
"pages": "118--128",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Y. Ajjour, W. Chen, J. Kiesel, H. Wachsmuth, and B. Stein. 2017. Unit segmentation of argumentative texts. In Proceedings of the 4th Workshop on Argument Mining, pages 118-128, Copenhagen, Denmark. Association for Computational Linguistics.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Modeling frames in argumentation",
"authors": [
{
"first": "Y",
"middle": [],
"last": "Ajjour",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Alshomary",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Wachsmuth",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Stein",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
"volume": "",
"issue": "",
"pages": "2922--2932",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Y. Ajjour, M. Alshomary, H. Wachsmuth, and B. Stein. 2019. Modeling frames in argumentation. In Proceed- ings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2922-2932, Hong Kong, China. Association for Computational Linguistics.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "A news editorial corpus for mining argumentation strategies",
"authors": [
{
"first": "K",
"middle": [],
"last": "Al-Khatib",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Wachsmuth",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Kiesel",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Hagen",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Stein",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers",
"volume": "",
"issue": "",
"pages": "3433--3443",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "K. Al-Khatib, H. Wachsmuth, J. Kiesel, M. Hagen, and B. Stein. 2016. A news editorial corpus for mining ar- gumentation strategies. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3433-3443, Osaka, Japan.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "De l'argumentation dans la langue\u00e0 la th\u00e9orie des topo\u00ef",
"authors": [
{
"first": "J",
"middle": [
"C"
],
"last": "Anscombre",
"suffix": ""
}
],
"year": 1995,
"venue": "La th\u00e9orie des topo\u00ef",
"volume": "",
"issue": "",
"pages": "11--47",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J.C. Anscombre. 1995. De l'argumentation dans la langue\u00e0 la th\u00e9orie des topo\u00ef. In La th\u00e9orie des topo\u00ef, pages 11-47. Kim\u00e9, Paris.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Annotation Guidelines for Labeling Topics, Aspects, Stance, Argumentativeness and Claims in Dutch social media comments, version 1.0",
"authors": [
{
"first": "N",
"middle": [],
"last": "Bauwelinck",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Lefever",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "3--15",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "N. Bauwelinck and E. Lefever. 2020. Annotation Guidelines for Labeling Topics, Aspects, Stance, Argumen- tativeness and Claims in Dutch social media comments, version 1.0. Technical report, Ghent University, LT3 15-01.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Identifying justifications in written dialogs",
"authors": [
{
"first": "O",
"middle": [],
"last": "Biran",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the 2011 IEEE Fifth International Conference on Semantic Computing, ICSC '11",
"volume": "",
"issue": "",
"pages": "162--168",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "O. Biran and O. Rambow. 2011. Identifying justifications in written dialogs. In Proceedings of the 2011 IEEE Fifth International Conference on Semantic Computing, ICSC '11, pages 162-168, USA. IEEE Computer So- ciety.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "A coefficient of agreement for nominal scales",
"authors": [
{
"first": "J",
"middle": [],
"last": "Cohen",
"suffix": ""
}
],
"year": 1960,
"venue": "Educational and Psychological Measurement",
"volume": "20",
"issue": "",
"pages": "37--46",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Cohen. 1960. A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20:37-46.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "i'm just saying",
"authors": [
{
"first": "R",
"middle": [
"T"
],
"last": "Craig",
"suffix": ""
},
{
"first": "A",
"middle": [
"L"
],
"last": "Sanusi",
"suffix": ""
}
],
"year": 2000,
"venue": "Discourse markers of standpoint continuity. Argumentation",
"volume": "14",
"issue": "",
"pages": "425--445",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R.T. Craig and A.L. Sanusi. 2000. 'i'm just saying...': Discourse markers of standpoint continuity. Argumentation, 14(4):425-445.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "What is the essence of a claim? crossdomain claim identification",
"authors": [
{
"first": "J",
"middle": [],
"last": "Daxenberger",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Eger",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Habernal",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Stab",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "2055--2066",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Daxenberger, S. Eger, I. Habernal, C. Stab, and I. Gurevych. 2017. What is the essence of a claim? cross- domain claim identification. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2055-2066, Copenhagen, Denmark. Association for Computational Linguistics.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Shell miner: Mining organizational phrases in argumentative texts in social media",
"authors": [
{
"first": "J",
"middle": [],
"last": "Du",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Jiang",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Song",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Liao",
"suffix": ""
}
],
"year": 2014,
"venue": "IEEE International Conference on Data Mining",
"volume": "",
"issue": "",
"pages": "797--802",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Du, J. Jiang, L. Yang, D. Song, and L. Liao. 2014. Shell miner: Mining organizational phrases in argumentative texts in social media. 2014 IEEE International Conference on Data Mining, pages 797-802.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Note sur l'argumentation et l'acte d'argumenter in concession et cons\u00e9cution dans le discours",
"authors": [
{
"first": "O",
"middle": [],
"last": "Ducrot",
"suffix": ""
}
],
"year": 1982,
"venue": "Cahiers de linguistique fran\u00e7aise",
"volume": "4",
"issue": "",
"pages": "143--163",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "O. Ducrot. 1982. Note sur l'argumentation et l'acte d'argumenter in concession et cons\u00e9cution dans le discours. In Cahiers de linguistique fran\u00e7aise, 4, pages 143-163, Gen\u00e8ve. Universit\u00e9 de Gen\u00e8ve.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "On the role of discourse markers for discriminating claims and premises in argumentative discourse",
"authors": [
{
"first": "J",
"middle": [],
"last": "Eckle-Kohler",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Kluge",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "2236--2242",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Eckle-Kohler, R. Kluge, and I. Gurevych. 2015. On the role of discourse markers for discriminating claims and premises in argumentative discourse. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2236-2242, Lisbon, Portugal. Association for Computational Linguistics.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Framing: Toward clarification of a fractured paradigm",
"authors": [
{
"first": "R",
"middle": [],
"last": "Entman",
"suffix": ""
}
],
"year": 1993,
"venue": "Journal of Communication",
"volume": "43",
"issue": "4",
"pages": "51--58",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R. Entman. 1993. Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4):51- 58.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Measuring nominal scale agreement among many raters",
"authors": [
{
"first": "J",
"middle": [
"L"
],
"last": "Fleiss",
"suffix": ""
}
],
"year": 1971,
"venue": "Psychological Bulletin",
"volume": "76",
"issue": "",
"pages": "378--382",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J.L. Fleiss. 1971. Measuring nominal scale agreement among many raters. Psychological Bulletin, 76:378-382.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Argumentation mining in user-generated web discourse",
"authors": [
{
"first": "I",
"middle": [],
"last": "Habernal",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2017,
"venue": "Computational Linguistics",
"volume": "43",
"issue": "1",
"pages": "125--179",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "I. Habernal and I. Gurevych. 2017. Argumentation mining in user-generated web discourse. Computational Linguistics, 43(1):125-179.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Bivariate agreement coefficients for reliability of data",
"authors": [
{
"first": "K",
"middle": [],
"last": "Krippendorff",
"suffix": ""
}
],
"year": 1970,
"venue": "Sociological methodology",
"volume": "",
"issue": "",
"pages": "139--150",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "K. Krippendorff. 1970. Bivariate agreement coefficients for reliability of data. Sociological methodology, pages 139-150.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Measuring the reliability of qualitative text analysis data. Quality quantity",
"authors": [
{
"first": "K",
"middle": [],
"last": "Krippendorff",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "38",
"issue": "",
"pages": "787--800",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "K. Krippendorff. 2004. Measuring the reliability of qualitative text analysis data. Quality quantity, 38:787-800.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Stance Detection: A Survey",
"authors": [
{
"first": "D",
"middle": [],
"last": "K\u00fc\u00e7\u00fck",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Can",
"suffix": ""
}
],
"year": 2020,
"venue": "ACM Computing Surveys",
"volume": "",
"issue": "1",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "D. K\u00fc\u00e7\u00fck and F. Can. 2020. Stance Detection: A Survey. ACM Computing Surveys, 53(1).",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "The measurement of observer agreement for categorical data",
"authors": [
{
"first": "J",
"middle": [
"R"
],
"last": "Landis",
"suffix": ""
},
{
"first": "G",
"middle": [
"G"
],
"last": "Koch",
"suffix": ""
}
],
"year": 1977,
"venue": "Biometrics",
"volume": "33",
"issue": "",
"pages": "159--174",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J.R. Landis and G.G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33:159-174.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Identifying high-level organizational elements in argumentative discourse",
"authors": [
{
"first": "N",
"middle": [],
"last": "Madnani",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Heilman",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Tetreault",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Chodorow",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "20--28",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "N. Madnani, M. Heilman, J. Tetreault, and M. Chodorow. 2012. Identifying high-level organizational elements in argumentative discourse. In Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 20-28, Montr\u00e9al, Canada. Association for Computational Linguistics.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Reader comments to online opinion journalism: A space of public deliberation",
"authors": [
{
"first": "E",
"middle": [],
"last": "Manosevitch",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Walker",
"suffix": ""
}
],
"year": 2009,
"venue": "10th International Symposium on Online Journalism",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "E. Manosevitch and D. Walker. 2009. Reader comments to online opinion journalism: A space of public delibera- tion. In 10th International Symposium on Online Journalism, Austin, Texas.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Extracting argument and domain words for identifying argument components in texts",
"authors": [
{
"first": "H",
"middle": [],
"last": "Nguyen",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Litman",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 2nd Workshop on Argumentation Mining",
"volume": "",
"issue": "",
"pages": "22--28",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "H. Nguyen and D. Litman. 2015. Extracting argument and domain words for identifying argument components in texts. In Proceedings of the 2nd Workshop on Argumentation Mining, pages 22-28, Denver, Colorado, USA. Association for Computational Linguistics.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Re-contextualising argumentative meanings: An adaptive perspective. Argumentation",
"authors": [
{
"first": "T",
"middle": [],
"last": "Nyan",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "31",
"issue": "",
"pages": "267--299",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "T. Nyan. 2017. Re-contextualising argumentative meanings: An adaptive perspective. Argumentation, 31:267- 299.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Identifying appropriate support for propositions in online user comments",
"authors": [
{
"first": "J",
"middle": [],
"last": "Park",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Cardie",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the First Workshop on Argumentation Mining",
"volume": "",
"issue": "",
"pages": "29--38",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Park and C. Cardie. 2014. Identifying appropriate support for propositions in online user comments. In Proceedings of the First Workshop on Argumentation Mining, pages 29-38, Baltimore, Maryland. Association for Computational Linguistics.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Measuring agreement on set-valued items (MASI) for semantic and pragmatic annotation",
"authors": [
{
"first": "R",
"middle": [],
"last": "Passonneau",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R. Passonneau. 2006. Measuring agreement on set-valued items (MASI) for semantic and pragmatic annota- tion. In Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06), Genoa, Italy. European Language Resources Association (ELRA).",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "An annotated corpus of argumentative microtexts",
"authors": [
{
"first": "A",
"middle": [],
"last": "Peldszus",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Stede",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the First Conference on Argumentation",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Peldszus and M. Stede. 2015. An annotated corpus of argumentative microtexts. In Proceedings of the First Conference on Argumentation, Lisbon, Portugal.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Deliberation 2.0: Comparing the deliberative quality of online news user comments across platforms",
"authors": [
{
"first": "I",
"middle": [],
"last": "Rowe",
"suffix": ""
}
],
"year": 2015,
"venue": "Journal of Broadcasting & Electronic Media",
"volume": "59",
"issue": "4",
"pages": "539--555",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "I. Rowe. 2015. Deliberation 2.0: Comparing the deliberative quality of online news user comments across plat- forms. Journal of Broadcasting & Electronic Media, 59(4):539-555.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Challenges of argument mining: Generating an argument synthesis based on the qualia structure",
"authors": [
{
"first": "P",
"middle": [],
"last": "Saint-Dizier",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of the 9th International Natural Language Generation conference",
"volume": "",
"issue": "",
"pages": "79--83",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "P. Saint-Dizier. 2016. Challenges of argument mining: Generating an argument synthesis based on the qualia structure. In Proceedings of the 9th International Natural Language Generation conference, pages 79-83, Edinburgh, UK. Association for Computational Linguistics.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Argumentation: Keeping Faith with Reason",
"authors": [
{
"first": "E",
"middle": [],
"last": "Schiappa",
"suffix": ""
},
{
"first": "J",
"middle": [
"P"
],
"last": "Nordin",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "E. Schiappa and J.P. Nordin. 2014. Argumentation: Keeping Faith with Reason. Pearson.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Discourse level opinion interpretation",
"authors": [
{
"first": "S",
"middle": [],
"last": "Somasundaran",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Wiebe",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Ruppenhofer",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of the 22nd International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "801--808",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "S. Somasundaran, J. Wiebe, and J. Ruppenhofer. 2008. Discourse level opinion interpretation. In Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), pages 801-808, Manchester, UK.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "Identifying argumentative discourse structures in persuasive essays",
"authors": [
{
"first": "C",
"middle": [],
"last": "Stab",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
"volume": "",
"issue": "",
"pages": "46--56",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "C. Stab and I. Gurevych. 2014. Identifying argumentative discourse structures in persuasive essays. In Proceed- ings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 46-56, Doha, Qatar. Association for Computational Linguistics.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "Coherence in political computer-mediated communication: analyzing topic relevance and drift in chat",
"authors": [
{
"first": "J",
"middle": [],
"last": "Stromer-Galley",
"suffix": ""
},
{
"first": "A",
"middle": [
"M"
],
"last": "Martinson",
"suffix": ""
}
],
"year": 2009,
"venue": "",
"volume": "3",
"issue": "",
"pages": "195--216",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Stromer-Galley and A.M. Martinson. 2009. Coherence in political computer-mediated communication: analyz- ing topic relevance and drift in chat. Discourse & Communication, 3(2):195-216.",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "From Connectives to Argumentative Markers: A Quest for Markers of Argumentative Moves and of Related Aspects of Argumentative Discourse",
"authors": [
{
"first": "A",
"middle": [],
"last": "Tseronis",
"suffix": ""
}
],
"year": 2011,
"venue": "",
"volume": "25",
"issue": "",
"pages": "427--447",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Tseronis. 2011. From Connectives to Argumentative Markers: A Quest for Markers of Argumentative Moves and of Related Aspects of Argumentative Discourse. Argumentation, 25(4):427-447.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "Strategic maneuvering: A synthetic recapitulation",
"authors": [
{
"first": "F",
"middle": [],
"last": "Van Eemeren",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Houtlosser",
"suffix": ""
}
],
"year": 2006,
"venue": "Argumentation",
"volume": "20",
"issue": "4",
"pages": "381--392",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "F. van Eemeren and P. Houtlosser. 2006. Strategic maneuvering: A synthetic recapitulation. Argumentation, 20(4):381-392-802.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "Some facets of argument mining for opinion analysis",
"authors": [
{
"first": "M",
"middle": [
"P G"
],
"last": "Villalba",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Saint-Dizier",
"suffix": ""
}
],
"year": 2012,
"venue": "COMMA",
"volume": "245",
"issue": "",
"pages": "23--34",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M.P.G. Villalba and P. Saint-Dizier. 2012. Some facets of argument mining for opinion analysis. COMMA, 245:23-34.",
"links": null
},
"BIBREF35": {
"ref_id": "b35",
"title": "Fact checking: Task definition and dataset construction",
"authors": [
{
"first": "A",
"middle": [],
"last": "Vlachos",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Riedel",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the ACL 2014 Workshop on Language Technologies and Computational Social Science",
"volume": "",
"issue": "",
"pages": "18--22",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Vlachos and S. Riedel. 2014. Fact checking: Task definition and dataset construction. In Proceedings of the ACL 2014 Workshop on Language Technologies and Computational Social Science, pages 18-22, Baltimore, MD, USA. Association for Computational Linguistics.",
"links": null
},
"BIBREF36": {
"ref_id": "b36",
"title": "Dialog theory for critical argumentation",
"authors": [
{
"first": "D",
"middle": [
"N"
],
"last": "Walton",
"suffix": ""
}
],
"year": 2007,
"venue": "John Benjamins",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "D.N. Walton. 2007. Dialog theory for critical argumentation. John Benjamins, Amsterdam.",
"links": null
},
"BIBREF37": {
"ref_id": "b37",
"title": "Stance-based Argument Mining -Modeling Implicit Argumentation Using Stance",
"authors": [
{
"first": "M",
"middle": [],
"last": "Wojatzki",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Zesch",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of the KONVENS",
"volume": "",
"issue": "",
"pages": "313--322",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. Wojatzki and T. Zesch. 2016. Stance-based Argument Mining -Modeling Implicit Argumentation Using Stance. In Proceedings of the KONVENS, pages 313-322.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"num": null,
"uris": null,
"text": "(4) 1[Feeling ill = stay at home!] 2[Easy!] 3[There are people who can't stand wearing the mask due to breathing problems!] 4[Wearing a mask in this heat is asking for trouble for those people.] 5[Keep your distance, sanitize your hands and stay AT HOME when ill.] 6[EASY...] 7.2 Segment-level annotation tasks (argumentativeness, claim detection)",
"type_str": "figure"
},
"TABREF1": {
"text": "Cohen's kappa agreement scores for pairs of annotators and Fleiss' kappa agreement for all three annotators for the tasks of argumentativeness and claim detection.",
"html": null,
"type_str": "table",
"content": "<table><tr><td>7 Analysis</td></tr><tr><td>7.1 Comment-level annotation tasks (topic, aspect, stance detection)</td></tr></table>",
"num": null
},
"TABREF2": {
"text": "But no matter what she says or does, it will never be enough.][...]",
"html": null,
"type_str": "table",
"content": "<table><tr><td>(5) 1[She's right.] 2[</td></tr></table>",
"num": null
}
}
}
} |