File size: 95,495 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 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 | {
"paper_id": "U07-1006",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T03:08:58.264491Z"
},
"title": "TAT: an author profiling tool with application to Arabic emails",
"authors": [
{
"first": "Dominique",
"middle": [],
"last": "Estival",
"suffix": "",
"affiliation": {},
"email": "destival@appen.com.au"
},
{
"first": "Tanja",
"middle": [],
"last": "Gaustad",
"suffix": "",
"affiliation": {},
"email": "tgaustad@appen.com.au"
},
{
"first": "Son",
"middle": [
"Bao"
],
"last": "Pham",
"suffix": "",
"affiliation": {},
"email": "sbpham@appen.com.au"
},
{
"first": "Will",
"middle": [],
"last": "Radford",
"suffix": "",
"affiliation": {},
"email": "wradford@appen.com.au"
},
{
"first": "Ben",
"middle": [],
"last": "Hutchinson",
"suffix": "",
"affiliation": {},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper reports on the application of the Text Attribution Tool (TAT) to profiling the authors of Arabic emails. The TAT system has been developed for the purpose of language-independent author profiling and has now been trained on two email corpora, English and Arabic. We describe the overall TAT system and the Machine Learning experiments resulting in classifiers for the different author traits. Predictions for demographic and psychometric author traits show improvements over the baseline for some of the author traits with both the English and the Arabic data. Arabic presents particular challenges for NLP and this paper describes more specifically the text processing components developed to handle Arabic emails.",
"pdf_parse": {
"paper_id": "U07-1006",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper reports on the application of the Text Attribution Tool (TAT) to profiling the authors of Arabic emails. The TAT system has been developed for the purpose of language-independent author profiling and has now been trained on two email corpora, English and Arabic. We describe the overall TAT system and the Machine Learning experiments resulting in classifiers for the different author traits. Predictions for demographic and psychometric author traits show improvements over the baseline for some of the author traits with both the English and the Arabic data. Arabic presents particular challenges for NLP and this paper describes more specifically the text processing components developed to handle Arabic emails.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The goal of the TAT project is to develop a languageindependent Text Attribution Tool (TAT) which can provide information on authors for a variety of document types and a range of languages. In the first implementation, the TAT has been developed for profiling the authors of email messages in English and Arabic, with other languages to be added in the future. (Estival et al., 2007) describes the Machine Learning experiments and the results obtained for the English email data. In this paper, we will focus on the results obtained for Arabic emails and will describe the aspects of the TAT which are particular to the processing of our Arabic email data.",
"cite_spans": [
{
"start": 362,
"end": 384,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "We first introduce the two tasks of author attribution and author profiling in Section 2. In Section 3, we describe the data set on which our tool was trained and tested. Section 4 contains an overall description of the TAT system, followed by a more specific description of the processing steps required for Arabic emails in Section 5. After presenting the general experimental setup in Section 6, we report on the results achieved for several demographic and psychometric traits in Section 7 and propose our general conclusions in Section 8.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The ability to recognise the identity or certain characteristics of authors automatically from texts has a number of potential applications. Author identification and author profiling can provide valuable information for marketing intelligence (Glance et al., 2005) , while the rapidly growing field of sentiment analysis and classification (Oberlander and Nowson, 2006 ) is another application where profiling can make a contribution. Also, author profiling forensics may be helpful in narrowing the choice of potential authors when identifying the source of a threat Argamon et al., 2005; Abbasi and Chen, 2005a) . Author attribution is the task of deciding for a given text which author, usually from a predefined set of authors, has written it. Historically, author identification has its roots in literature, with studies of the Bible (Friedmann, 1997) , Shakespeare (Ledger and Merriam, 1994) or the Federalist Pa-pers (Mosteller and Wallace, 1964) . Recently, author identification has also been applied to more informal texts, such as emails (de Vel, 2000; de Vel et al., 2001; , newsgroup messages (Zheng et al., 2003; Zheng et al., 2006) or blog entries Oberlander and Nowson, 2006) . Author profiling is the task of predicting one or more traits for the author of a text and the author profile consists of the set of traits predicted for that author. A major difference between author profiling and author attribution is that it is possible to predict author traits even when the training data does not contain any document by the actual author. Another difference is that greater accuracy can be expected for author profiling when the training data is made up of more authors, because the models for each trait are expected to be more robust. The accuracy of author identification, on the other hand, can be expected to decrease if the number of potential authors is increased.",
"cite_spans": [
{
"start": 244,
"end": 265,
"text": "(Glance et al., 2005)",
"ref_id": "BIBREF23"
},
{
"start": 341,
"end": 369,
"text": "(Oberlander and Nowson, 2006",
"ref_id": "BIBREF30"
},
{
"start": 569,
"end": 590,
"text": "Argamon et al., 2005;",
"ref_id": "BIBREF4"
},
{
"start": 591,
"end": 614,
"text": "Abbasi and Chen, 2005a)",
"ref_id": "BIBREF0"
},
{
"start": 840,
"end": 857,
"text": "(Friedmann, 1997)",
"ref_id": "BIBREF22"
},
{
"start": 872,
"end": 898,
"text": "(Ledger and Merriam, 1994)",
"ref_id": "BIBREF27"
},
{
"start": 925,
"end": 954,
"text": "(Mosteller and Wallace, 1964)",
"ref_id": "BIBREF28"
},
{
"start": 1050,
"end": 1064,
"text": "(de Vel, 2000;",
"ref_id": "BIBREF17"
},
{
"start": 1065,
"end": 1085,
"text": "de Vel et al., 2001;",
"ref_id": "BIBREF15"
},
{
"start": 1107,
"end": 1127,
"text": "(Zheng et al., 2003;",
"ref_id": "BIBREF33"
},
{
"start": 1128,
"end": 1147,
"text": "Zheng et al., 2006)",
"ref_id": "BIBREF34"
},
{
"start": 1164,
"end": 1192,
"text": "Oberlander and Nowson, 2006)",
"ref_id": "BIBREF30"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Author attribution and author profiling",
"sec_num": "2"
},
{
"text": "In (Estival et al., 2007) , we presented our work on author profiling for English emails and discussed the literature on previous research in that area for English texts. For Arabic texts, very few studies have been published in the area of author attribution; the only work we know of can be found in (Abbasi and Chen, 2005b) and (Abbasi and Chen, 2005a) . We are not aware of any work on author profiling for Arabic.",
"cite_spans": [
{
"start": 3,
"end": 25,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
},
{
"start": 302,
"end": 326,
"text": "(Abbasi and Chen, 2005b)",
"ref_id": "BIBREF1"
},
{
"start": 331,
"end": 355,
"text": "(Abbasi and Chen, 2005a)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Author attribution and author profiling",
"sec_num": "2"
},
{
"text": "For author attribution in Arabic, Abbasi and Chen (2005a) apply two different machine learners to a dataset of Arabic forum messages by 20 authors (with 20 messsages per author). The feature set comprises lexical, syntactic, structural and contentspecific features, including a number of features specifically tailored to Arabic. These relate mostly to inflection (counting roots rather than inflected words), word length (adjusting the range for the Arabic word length distribution to reflect the fact that Arabic words tend to be shorter than English ones) and elongation dashes (excluding them from word length measurements, but tracking their usage). The main conclusion reached is that using an SVM classifier with all the features achieves the best accuracy on their data set, but that the overall performance is lower than for English. Abbasi and Chen (2005b) use the same approach, and in addition include an in-depth comparison between the feature sets and results for English and Arabic on forum messages.",
"cite_spans": [
{
"start": 34,
"end": 57,
"text": "Abbasi and Chen (2005a)",
"ref_id": "BIBREF0"
},
{
"start": 843,
"end": 866,
"text": "Abbasi and Chen (2005b)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Author attribution and author profiling",
"sec_num": "2"
},
{
"text": "The data collected for this project and used for training the TAT consists of two sets of emails from 1033 English speakers and from 1030 Arabic speakers. The English data set contains emails written by both native and non-native speakers of English: native speakers of US English, native speakers of Australian/NZ English, native speakers of Spanish from the US and native speakers of Egyptian Arabic from Egypt. The Arabic data set contains emails written by native speakers of Egyptian Arabic, as described in more detail below. In the rest of this section, we focus on the data collection and validation processes for Arabic emails.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "For the Arabic email data set, the collection was conducted in Egypt and all the writers were native speakers of Egyptian Arabic. Compared with the English email data set, a special feature is the encoding of the input for Arabic email. The widespread use of the Internet and even more of text messages via mobile phones without the possibility of Arabic script input has led to the use of the Latin alphabet and the development of some transliteration conventions in the Arabic speaking world. Even though Arabic keyboards are now more common, people still sometimes write email using a Latin keyboard. However, there are no strict rules for the conversion of Arabic script into Latin characters and the way of writing emails with a Latin keyboard varies greatly according to dialect or country, and even across individuals. Table 1 gives an overview of the number of authors, number of emails and number of words for both the Arabic and English data set. We also include the proportion of emails in Arabic script and in Latin script for the Arabic data.",
"cite_spans": [],
"ref_spans": [
{
"start": 826,
"end": 833,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Data Collection",
"sec_num": "3.1"
},
{
"text": "The data collection process for the Arabic data differed slightly from the process described in (Estival et al., 2007) for the collection of the English data, in that the respondents had to come to a central collection location. Nevertheless, the recruitment pro- cess also included notification of privacy and the assurance that their identity would be protected. The respondents agreed to fill out a web questionnaire to provide demographic and psychometric information about themselves and to donate at least ten email messages. Demographic traits cover basic demographic information about the author, namely age, gender, and level of education. The psychometric traits of the Arabic collection are based on a customized version of the short Eysenck Personality Questionnaire Revised (EPQR-S) (Francis et al., 2006) , consisting of 51 questions. These questions aim to identify four psychometric traits: extraversion, lie (or social desirability), neuroticism (or emotionality), and psychoticism (or toughmindedness).",
"cite_spans": [
{
"start": 96,
"end": 118,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
},
{
"start": 796,
"end": 818,
"text": "(Francis et al., 2006)",
"ref_id": "BIBREF20"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Collection",
"sec_num": "3.1"
},
{
"text": "After completing the questionnaire, the writers either composed new email messages which they then sent to their recipients and also forwarded to the data collection email address, or directly forwarded previously sent emails, e.g. from their email client \"SentBox\".",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Collection",
"sec_num": "3.1"
},
{
"text": "We collected at least 10 emails from each author, for a total number of 2000 words per author in the Arabic data set. Research has shown that the more complex morphology of Arabic (combined with a rich vocabulary) leads to a higher degree of inherent sparseness in Arabic data compared to similar English data. This suggests that larger amounts of data are needed for statistical Natural Language Processing (NLP) applications in Arabic (Goweder and Roeck, 2001) . Therefore, while the minimum was set at 1000 words per writer for the English data, it was 2000 words per writer for the Arabic data.",
"cite_spans": [
{
"start": 437,
"end": 462,
"text": "(Goweder and Roeck, 2001)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Collection",
"sec_num": "3.1"
},
{
"text": "The email messages were checked manually to filter out erroneous content such as foreign language emails or forwarded chain letters and to ensure con-sistency and accuracy of the documents in the corpus. As with any data collection of email, plagiarism and copying were issues that required careful checking of all the data received and we developed a plagiarism detecter to reject emails which had already been submitted.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Validation",
"sec_num": "3.2"
},
{
"text": "For both the English and the Arabic data collections, a minimum number of 5 lines per email had been set. Because Arabic writers often do not use new lines or new paragraphs, this had to be measured visually on the screen for the Arabic data.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Validation",
"sec_num": "3.2"
},
{
"text": "In summary, the final Arabic data set contains a combined total of 8028 email messages, from 1030 writers who met the following criteria: 1) a valid questionnaire was received for each author; 2) there are at least 5 valid email messages for the author; and 3) the total word count for that author's valid email messages is at least 2000 words. Figure 1 gives a high-level overview of the TAT system. The system includes several data repositories and a number of components for deriving features and for building classifiers. While the current focus is on processing email input in English and in Arabic, the underlying processing architecture is language independent and will be extended to other types of documents and to other languages.",
"cite_spans": [],
"ref_spans": [
{
"start": 345,
"end": 353,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Data Validation",
"sec_num": "3.2"
},
{
"text": "The modular processing architecture is organized around a chain of processing modules. This chain allows flexible experimentation with various processing modules. At the same time, it provides a robust software framework that promotes reuse and supports flexible deployment options by connecting specific modules together for the task at hand.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "System Description",
"sec_num": "4"
},
{
"text": "Each processing module consumes objects from its input, such as documents, and emits objects containing the analysis of the input objects. The analysis of a document is represented in stand-off anno- The process is data-driven in the sense that the output of each processing module depends on its input rather than on the way the module is combined with other modules in the chain. This enables processing modules to be reused in different processing chains and in different control environments as long as the input requirements are met.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "System Description",
"sec_num": "4"
},
{
"text": "After the initial data collection and validation, several processing steps are needed, including character encoding normalisation, document structure parsing, text processing and linguistic analysis. The results of this processing provides the input to the feature extraction phase. A more extensive description of the processing steps for the English data can be found in (Estival et al., 2007) .",
"cite_spans": [
{
"start": 373,
"end": 395,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "Arabic emails present a number of challenges for NLP, including different ways of writing Arabic in Latin script (so-called \"franco-arabic\"); typical spelling variants in the Egyptian dialect of Arabic and possible spelling normalisation; morphol-ogy; English loanwords, spelling errors, and typos.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "The Arabic processing modules perform the following functions:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "1. Language and character-set identification. The language is tagged as either English or Arabic and the script as either \"roman\" or \"arabic\", using character-based language models.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "2. Document structure parsing. This stage distinguishes the text of the email written by the author from other types of elements (adverts, signatures, reply lines, or quoted text). 3. Tokenisation. The input text is split into paragraphs, sentences and words. Words are mainly strings of alphanumeric characters, with a few major exceptions for transliterated Arabic: some two character sequences, e.g. \" 3' \", can indicate a single Arabic glyph, namely .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "4. Character set normalisation. In order to achieve a normalised, unambiguous input for Arabic and Latin scripted texts, words are converted to ASCII only characters using the Buckwalter transliteration scheme (Buckwalter, 2000; Buckwalter, 2002) . The Buckwalter scheme is very commonly used in NLP for Arabic because it represents Arabic orthography strictly on a one-to-one basis (unlike common romanization schemes that add morphological information not actually expressed in the Arabic script).",
"cite_spans": [
{
"start": 210,
"end": 228,
"text": "(Buckwalter, 2000;",
"ref_id": "BIBREF8"
},
{
"start": 229,
"end": 246,
"text": "Buckwalter, 2002)",
"ref_id": "BIBREF9"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Arabic Text processing",
"sec_num": "5"
},
{
"text": "Informal written Arabic often contains non-standard spellings (Buckwalter, 2004; Goweder and Roeck, 2001) . We have therefore normalised the spelling based on common spelling variations in Egyptian Arabic. Examples include word final becoming or word final becoming . Also, depending on the context, can take different chairs or appear by itself on the line.",
"cite_spans": [
{
"start": 62,
"end": 80,
"text": "(Buckwalter, 2004;",
"ref_id": "BIBREF10"
},
{
"start": 81,
"end": 105,
"text": "Goweder and Roeck, 2001)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Spelling normalisation.",
"sec_num": "5."
},
{
"text": "We normalise the hamza chair to .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Spelling normalisation.",
"sec_num": "5."
},
{
"text": "6. Morphological analysis. By comparing the normalised version of a given word with dictionaries of prefixes, suffixes and clitics, linguistic features such as number and person are added, and the remainder of the word is tagged as a stem. The root letters of the stem are then predicted using simple linguistic heuristics, e.g. long vowels are less likely to be root letters.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Spelling normalisation.",
"sec_num": "5."
},
{
"text": "7. Lexicon taggers. The following word classes are currently tagged: conjunctions, prepositions, pronouns, discourse particles, interrogative pronouns, English loan words, colloquial Egyptian words, and frequent roots.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Spelling normalisation.",
"sec_num": "5."
},
{
"text": "8. Named Entity Recognition. Named entities which are not language-dependent are tagged. These include numeric dates, numeric times, phone numbers, email addresses and URLs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Spelling normalisation.",
"sec_num": "5."
},
{
"text": "Recent years have seen an exponential increase in the use of statistical language processing techniques for a wide range of tasks. In particular, text and document classification problems greatly benefit from statistical approaches. Author profiling can be viewed as a type of document classification task, where the classes correspond to traits of the authors. These traits are arranged along various dimensions and the different options for each dimension are mutually exclusive. For example \"male\" and \"female\" are the possibilities for the gender dimension. For each dimension, the email and questionnaire data are used to construct classifiers, using a range of Machine Learning (ML) techniques.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental setup",
"sec_num": "6"
},
{
"text": "Each document constitutes a single data instance for the purposes of the experiments. For each experiment, ten-fold cross-validation was used, so the results reported in Section 7 are on the entire data set. Once the best combination of ML classifiers, parameters and feature selection has been determined during development, that model is used to classify the test data to evaluate the performance for a given trait.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental setup",
"sec_num": "6"
},
{
"text": "We distinguish three different demographic and four psychometric traits in the experiments presented in this paper, namely age, gender, and level of education for the demographic traits, and extraversion, lie, neuroticism, and psychoticism for the psychometric traits (Eysenck and Eysenck, 1975) . This information is extracted from the questionnaires.",
"cite_spans": [
{
"start": 268,
"end": 295,
"text": "(Eysenck and Eysenck, 1975)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Traits and classes",
"sec_num": "6.1"
},
{
"text": "For the traits taking numerical values, subjects were split into three classes based on the first and third quartiles. Table 2 summarizes the data distribution for each trait across these classes.",
"cite_spans": [],
"ref_spans": [
{
"start": 119,
"end": 126,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Traits and classes",
"sec_num": "6.1"
},
{
"text": "A major difference with regard to the psychometric traits described in (Estival et al., 2007) lies in the fact that we used two different questionnaires for English and for Arabic. The International Personality Item Pool questionnaire (IPIP) (Buchanan et al., 2005) , which was used for collecting the English data, yields five psychometric traits: agreeableness, extraversion, neuroticism, conscientiousness, and openness (also referred to as the \"Big Five\") (Norman, 1963) . However the questionnaire for the Arabic data collection was based on the short form of the revised Eysenck Personality Questionnaire, the EPQR-S first developed for a study in Germany (Francis et al., 2006) , and the adaptation of the EPQ for Arabic (Abdel-Khalek and Eysenenck, 1983) . The EPQ (Eysenck and Eysenck, 1975) aims to analyze personality along four traits, namely extraversion, neuroticism, psychoticism, and lie. This entails that the Arabic results for the psychometric traits are not directly comparable to the English ones.",
"cite_spans": [
{
"start": 71,
"end": 93,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
},
{
"start": 242,
"end": 265,
"text": "(Buchanan et al., 2005)",
"ref_id": "BIBREF7"
},
{
"start": 460,
"end": 474,
"text": "(Norman, 1963)",
"ref_id": "BIBREF29"
},
{
"start": 662,
"end": 684,
"text": "(Francis et al., 2006)",
"ref_id": "BIBREF20"
},
{
"start": 728,
"end": 762,
"text": "(Abdel-Khalek and Eysenenck, 1983)",
"ref_id": "BIBREF2"
},
{
"start": 773,
"end": 800,
"text": "(Eysenck and Eysenck, 1975)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Traits and classes",
"sec_num": "6.1"
},
{
"text": "Gender: Level of education: <25 691Male ( Table 2 : Traits and Classes, with frequencies in parentheses where applicable.",
"cite_spans": [],
"ref_spans": [
{
"start": 42,
"end": 49,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Age:",
"sec_num": null
},
{
"text": "For each document, a feature vector is calculated. A feature is typically a descriptive statistic calculated from both the raw text and the annotations. For example, a feature might express the relative frequency of two different annotation types (e.g. number of words/number of sentences), or the presence or absence of an annotation type (e.g. Advert).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Features",
"sec_num": "6.2"
},
{
"text": "For the Arabic data, 518 features were calculated. These were divided into several subgroups shown in Table 3 . The main purpose of the groupings was to make more informed choices during the feature selection stage and to facilitate experimentation with various combinations of feature groups.",
"cite_spans": [],
"ref_spans": [
{
"start": 102,
"end": 109,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Features",
"sec_num": "6.2"
},
{
"text": "Character-level features cover features such as the frequency of punctuation characters or word length. Arabic-specific character features include information on spelling normalisation and Arabic special characters. Morphological features mainly encode information on number, person and gender markers, such as clitics or suffixes. Lexical features include certain POS tags (e.g. preposition) and whether a word is an English loanword or specific to the Egyptian dialect.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Features",
"sec_num": "6.2"
},
{
"text": "Classifiers for predicting author traits from the linguistic features were trained using the WEKA toolkit (Witten and Frank, 2005) . During training, classifiers are created by the selection of sets of features for each trait, and classifier parameters are tuned through cross-validation. To evaluate and test the classifiers, new documents are given as input and existing classifiers are selected to predict author traits.",
"cite_spans": [
{
"start": 106,
"end": 130,
"text": "(Witten and Frank, 2005)",
"ref_id": "BIBREF32"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Classification algorithms and feature selection",
"sec_num": "6.3"
},
{
"text": "The machine learning algorithms tested include decision trees (J48 (Quinlan, 1993) , RandomForest (Breiman, 2001) ), lazy learners (IBk (Aha et al., 1991) ), rule-based learners (JRip (Cohen, 1995) ), Support Vector Machines (SMO (Keerthi et al., 2001 )), as well as ensemble/meta-learners (Bagging (Breiman, 1996) , AdaBoostM1 (Freund and Schapire, 1996) ). These algorithms were used in combination with feature selection methods based on either a feature subset evaluator together with a search method (consistency subset evaluator with a best-first search) or a single attribute evaluator with various numbers of attributes selected (\u03c7 2 , GainRatio, and InformationGain) (see chapter 10.8 in (Witten and Frank, 2005) for details).",
"cite_spans": [
{
"start": 67,
"end": 82,
"text": "(Quinlan, 1993)",
"ref_id": "BIBREF31"
},
{
"start": 98,
"end": 113,
"text": "(Breiman, 2001)",
"ref_id": "BIBREF6"
},
{
"start": 136,
"end": 154,
"text": "(Aha et al., 1991)",
"ref_id": "BIBREF3"
},
{
"start": 184,
"end": 197,
"text": "(Cohen, 1995)",
"ref_id": "BIBREF11"
},
{
"start": 230,
"end": 251,
"text": "(Keerthi et al., 2001",
"ref_id": "BIBREF25"
},
{
"start": 299,
"end": 314,
"text": "(Breiman, 1996)",
"ref_id": "BIBREF5"
},
{
"start": 328,
"end": 355,
"text": "(Freund and Schapire, 1996)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Classification algorithms and feature selection",
"sec_num": "6.3"
},
{
"text": "The results shown here were computed on the Arabic email data set described in Section 3 using the different classifiers and general setup introduced in Section 6. Table 4 shows the results on all seven traits (demographic and psychometric), including the respective baseline associated with each separate classification task. We also state which settings (ML algorithm, feature set, and feature selection) were used to achieve the results reported. Education and gender are both binary classification tasks, whereas age uses three classes. All the psychometric traits are divided into three classes (see Section 6.1 and Table 2 for details on the exact split).",
"cite_spans": [],
"ref_spans": [
{
"start": 164,
"end": 171,
"text": "Table 4",
"ref_id": null
},
{
"start": 621,
"end": 628,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "The results show that for six out of the seven traits Table 4 : Results for all demographic and psychometric traits on Arabic email data.",
"cite_spans": [],
"ref_spans": [
{
"start": 54,
"end": 61,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "tested for, classification is significantly 1 improved over the baseline. For education, virtually no improvement can be seen which is due to the extremely skewed data (as indicated by the very high baseline of 93.62%). Even though the baselines for the other demographic traits are also quite high, our system still achieves a better classification accuracy for age and gender than the majority baseline. The better result for gender can in part be explained by the fact that gender is morphologically marked in Arabic. One of the relevant constructions with respect to identifying an author's gender are predicative sentences with first person subjects. For example, in the Arabic equivalent of \"I am happy\", happy is morphologically marked as either feminine or masculine. Since our features include morphological information, our classifier detects gender differences very accurately. A more detailed analysis of the effects of each feature group on the prediction of gender (shown in Table 5 ) reveals that lexical features are also of great assistance. Table 6 shows the results for English and Arabic demographic traits that are directly comparable. This seems to confirm previous results showing that predicting author traits for Arabic is likely to be 1 Significance was tested using a \u03c7 2 test with p=0.01. more difficult than for English. One should not forget, however, that the baselines for all Arabic demographic traits are extremely high which means that little data is available for the minority classes.",
"cite_spans": [],
"ref_spans": [
{
"start": 989,
"end": 996,
"text": "Table 5",
"ref_id": "TABREF4"
},
{
"start": 1059,
"end": 1066,
"text": "Table 6",
"ref_id": "TABREF5"
}
],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "For the psychometric traits, we achieve similar improvements over the baseline as for English. This is particularly encouraging, as most research on Arabic author attribution has shown results for Arabic to be lower than for English. It seems that predicting a profile rather than an identity might be advantageous for Arabic, or at least a viable back-off option.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "From our results, two ML algorithms emerge as best performing for all traits examined, namely SMO and Bagging. Bagging seems to profit from feature selection whereas the Support Vector Machine based SMO algorithm does not show additional improvements when combined with feature selection. This differs slightly from our results with the English data set, where no clear conclusion could be drawn with regard to the usefulness of feature selection for different algorithms.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "An analysis of the results shows that the highest accuracy is achieved by including all available features, with the exclusion of a single feature group for age (arabicLexical) and for extraversion (arabic-Morphological). ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results and discussion",
"sec_num": "7"
},
{
"text": "We have presented some results of experiments to automatically predict author traits from email messages. This work is of interest for a number of potential applications, from marketing intelligence to sentiment analysis. The results presented in this paper were conducted on the Arabic subset of the email data we have collected (containing approximately 8028 emails). The experiments reported here were aimed at discovering how well a range of ML algorithms perform on our data set for three demographic and four psychometric author traits. Our results support the conclusions drawn in (Estival et al., 2007) that the chosen approach works well for author profiling and that using different classifiers in combination with a subset of available features can be beneficial for predicting single traits.",
"cite_spans": [
{
"start": 588,
"end": 610,
"text": "(Estival et al., 2007)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and future work",
"sec_num": "8"
},
{
"text": "Future work will include the extension of the TAT to other document types and other languages. For Arabic text processing, it might be fruitful to investigate a more sophisticated analysis of words into roots (Darwish, 2002; de Roeck and Al-Fares, 2000) .",
"cite_spans": [
{
"start": 209,
"end": 224,
"text": "(Darwish, 2002;",
"ref_id": "BIBREF13"
},
{
"start": 225,
"end": 253,
"text": "de Roeck and Al-Fares, 2000)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and future work",
"sec_num": "8"
}
],
"back_matter": [
{
"text": "This research was carried out within the framework of a BAA grant (IS-QD-2467). We would like to thank Judith Bishop for her important contribution to the project in successfully managing the data collection.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": null
},
{
"text": "Description all all features for Arabic arabicNamedEntities language independent named entities, such as URLs arabicChar Arabic character features arabicMorphological Arabic morphology features arabicLexical Arabic lexicon and word features ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Feature group",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Applying authorship analysis to Arabic web content",
"authors": [
{
"first": "Ahmed",
"middle": [],
"last": "Abbasi",
"suffix": ""
},
{
"first": "Hsinchun",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2005,
"venue": "Intelligence and Security Informatics, Proceedings of the IEEE International Conference on Intelligence and Security Informatics (ISI 2005)",
"volume": "",
"issue": "",
"pages": "183--197",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ahmed Abbasi and Hsinchun Chen. 2005a. Applying authorship analysis to Arabic web content. In Paul B. Kantor et al., editor, Intelligence and Security In- formatics, Proceedings of the IEEE International Con- ference on Intelligence and Security Informatics (ISI 2005), pages 183-197. Springer.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Applying authorship analysis to extremist-group web forum messages",
"authors": [
{
"first": "Ahmed",
"middle": [],
"last": "Abbasi",
"suffix": ""
},
{
"first": "Hsinchun",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2005,
"venue": "IEEE Intelligent Systems",
"volume": "20",
"issue": "5",
"pages": "67--75",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ahmed Abbasi and Hsinchun Chen. 2005b. Apply- ing authorship analysis to extremist-group web forum messages. IEEE Intelligent Systems, 20(5):67-75.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "A cross-cultural study of personality: Egypt and England",
"authors": [
{
"first": "Ahmed",
"middle": [],
"last": "Abdel",
"suffix": ""
},
{
"first": "-",
"middle": [],
"last": "Khalek",
"suffix": ""
},
{
"first": "Sybil",
"middle": [],
"last": "Eysenenck",
"suffix": ""
}
],
"year": 1983,
"venue": "Research in Behaviour and Personality",
"volume": "3",
"issue": "",
"pages": "215--226",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ahmed Abdel-Khalek and Sybil Eysenenck. 1983. A cross-cultural study of personality: Egypt and Eng- land. Research in Behaviour and Personality, 3:215- 226.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Instance-based learning algorithms",
"authors": [
{
"first": "David",
"middle": [],
"last": "Aha",
"suffix": ""
},
{
"first": "Dennis",
"middle": [],
"last": "Kibler",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Albert",
"suffix": ""
}
],
"year": 1991,
"venue": "Machine Learning",
"volume": "6",
"issue": "1",
"pages": "37--66",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "David Aha, Dennis Kibler, and Mark Albert. 1991. Instance-based learning algorithms. Machine Learn- ing, 6(1):37-66.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Lexical predictors of personality type",
"authors": [
{
"first": "Shlomo",
"middle": [],
"last": "Argamon",
"suffix": ""
},
{
"first": "Sushant",
"middle": [],
"last": "Dhawle",
"suffix": ""
},
{
"first": "Mosche",
"middle": [],
"last": "Koppel",
"suffix": ""
},
{
"first": "James",
"middle": [
"W"
],
"last": "Pennebaker",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the 2005 Joint Annual Meeting of the Interface and the Classification Society of North America",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Shlomo Argamon, Sushant Dhawle, Mosche Koppel, and James W. Pennebaker. 2005. Lexical predictors of personality type. In Proceedings of the 2005 Joint An- nual Meeting of the Interface and the Classification Society of North America, St. Louis.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Bagging predictors",
"authors": [
{
"first": "Leo",
"middle": [],
"last": "Breiman",
"suffix": ""
}
],
"year": 1996,
"venue": "Machine Learning",
"volume": "24",
"issue": "",
"pages": "123--140",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Leo Breiman. 1996. Bagging predictors. Machine Learning, 24(2):123-140.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Random forests. Machine Learning",
"authors": [
{
"first": "Leo",
"middle": [],
"last": "Breiman",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "45",
"issue": "",
"pages": "5--32",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Leo Breiman. 2001. Random forests. Machine Learn- ing, 45(1):5-32.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Implementing a Five-Factor personality inventory for use on the internet",
"authors": [
{
"first": "Tom",
"middle": [],
"last": "Buchanan",
"suffix": ""
},
{
"first": "John",
"middle": [
"A"
],
"last": "Johnson",
"suffix": ""
},
{
"first": "Lewis",
"middle": [
"R"
],
"last": "Goldberg",
"suffix": ""
}
],
"year": 2005,
"venue": "European Journal of Psychological Assessment",
"volume": "21",
"issue": "",
"pages": "115--127",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tom Buchanan, John A. Johnson, and Lewis R. Gold- berg. 2005. Implementing a Five-Factor personality inventory for use on the internet. European Journal of Psychological Assessment, 21:115-127.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Arabic transliteration",
"authors": [
{
"first": "Tim",
"middle": [],
"last": "Buckwalter",
"suffix": ""
}
],
"year": 2000,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tim Buckwalter. 2000. Arabic transliteration.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Arabic morphological analyzer",
"authors": [
{
"first": "Tim",
"middle": [],
"last": "Buckwalter",
"suffix": ""
}
],
"year": 2002,
"venue": "Linguistic Data Consortium (LDC2002L49)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tim Buckwalter. 2002. Arabic morphological analyzer. Linguistic Data Consortium (LDC2002L49).",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Issues in Arabic orthography and morphology analysis",
"authors": [
{
"first": "Tim",
"middle": [],
"last": "Buckwalter",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the COLING 2004 Workshop on computational approaches to Arabic script-based languages",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tim Buckwalter. 2004. Issues in Arabic orthography and morphology analysis. In Proceedings of the COLING 2004 Workshop on computational approaches to Ara- bic script-based languages, Geneva.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Fast effective rule induction",
"authors": [
{
"first": "William",
"middle": [],
"last": "Cohen",
"suffix": ""
}
],
"year": 1995,
"venue": "Twelfth International Conference on Machine Learning",
"volume": "",
"issue": "",
"pages": "115--123",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "William Cohen. 1995. Fast effective rule induction. In Twelfth International Conference on Machine Learn- ing, pages 115-123.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Gender-preferential text mining of e-mail discourse",
"authors": [
{
"first": "Malcolm",
"middle": [],
"last": "Corney",
"suffix": ""
},
{
"first": "Alison",
"middle": [],
"last": "Olivier De Vel",
"suffix": ""
},
{
"first": "George",
"middle": [],
"last": "Anderson",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mohay",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 18th Annual Computer Security Applications Conference (AC-SAC 2002)",
"volume": "",
"issue": "",
"pages": "282--292",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Malcolm Corney, Olivier de Vel, Alison Anderson, and George Mohay. 2002. Gender-preferential text min- ing of e-mail discourse. In Proceedings of the 18th An- nual Computer Security Applications Conference (AC- SAC 2002), pages 282-292, Las Vegas.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Building a shallow Arabic morphological analyzer in one day",
"authors": [
{
"first": "Kareem",
"middle": [],
"last": "Darwish",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of ACL 2002 Workshop on computational approaches to Semitic languages",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kareem Darwish. 2002. Building a shallow Arabic mor- phological analyzer in one day. In Proceedings of ACL 2002 Workshop on computational approaches to Semitic languages, Philadelphia.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "A morphologically sensitive clustering algorithm for identifying Arabic roots",
"authors": [],
"year": 2000,
"venue": "Proceedings of ACL 2000",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Anne de Roeck and Waleed Al-Fares. 2000. A morpho- logically sensitive clustering algorithm for identifying Arabic roots. In Proceedings of ACL 2000, Toulouse.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Mining email content for author identification forensics",
"authors": [
{
"first": "Alison",
"middle": [],
"last": "Olivier De Vel",
"suffix": ""
},
{
"first": "Malcolm",
"middle": [],
"last": "Anderson",
"suffix": ""
},
{
"first": "George",
"middle": [],
"last": "Corney",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mohay",
"suffix": ""
}
],
"year": 2001,
"venue": "SIGMOD Record",
"volume": "30",
"issue": "4",
"pages": "55--64",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Olivier de Vel, Alison Anderson, Malcolm Corney, and George Mohay. 2001. Mining email content for author identification forensics. SIGMOD Record, 30(4):55-64.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "E-mail authorship attribution for computer forensics",
"authors": [
{
"first": "Alison",
"middle": [],
"last": "Olivier De Vel",
"suffix": ""
},
{
"first": "Malcolm",
"middle": [],
"last": "Anderson",
"suffix": ""
},
{
"first": "George",
"middle": [],
"last": "Corney",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mohay",
"suffix": ""
}
],
"year": 2002,
"venue": "Data Mining for Security Applications",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Olivier de Vel, Alison Anderson, Malcolm Corney, and George Mohay. 2002. E-mail authorship attribution for computer forensics. In Daniel Barbara and Sushil Jajodia, editors, Data Mining for Security Applica- tions. Kluwer Academic Publishers.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Mining e-mail authorship",
"authors": [
{
"first": "Vel",
"middle": [],
"last": "Olivier De",
"suffix": ""
}
],
"year": 2000,
"venue": "Proceedings of the Workshop on Text Mining, ACM International Conference on Knowledge Discovery and Data Mining (KDD'2000)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Olivier de Vel. 2000. Mining e-mail authorship. In Pro- ceedings of the Workshop on Text Mining, ACM In- ternational Conference on Knowledge Discovery and Data Mining (KDD'2000), Boston.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Author profiling for English emails",
"authors": [
{
"first": "Dominique",
"middle": [],
"last": "Estival",
"suffix": ""
},
{
"first": "Tanja",
"middle": [],
"last": "Gaustad",
"suffix": ""
},
{
"first": "Son",
"middle": [
"Bao"
],
"last": "Pham",
"suffix": ""
},
{
"first": "Will",
"middle": [],
"last": "Radford",
"suffix": ""
},
{
"first": "Ben",
"middle": [],
"last": "Hutchinson",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 10th Conference of the Pacific Association for Computational Linguistics (PACLING 2007), pages ??-??",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dominique Estival, Tanja Gaustad, Son Bao Pham, Will Radford, and Ben Hutchinson. 2007. Author profiling for English emails. In Proceedings of the 10th Confer- ence of the Pacific Association for Computational Lin- guistics (PACLING 2007), pages ??-??, Melbourne.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Manual of the Eysenck Personality Questionnaire. Hooder and Stoughton Educational",
"authors": [
{
"first": "Hans",
"middle": [],
"last": "Eysenck",
"suffix": ""
},
{
"first": "Sybil",
"middle": [],
"last": "Eysenck",
"suffix": ""
}
],
"year": 1975,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hans Eysenck and Sybil Eysenck. 1975. Manual of the Eysenck Personality Questionnaire. Hooder and Stoughton Educational, London.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "The short-form revised Eysenenck personality questionnaire (EPQR-S): A German edition",
"authors": [
{
"first": "Leslie",
"middle": [],
"last": "Francis",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Lewis",
"suffix": ""
},
{
"first": "Hans-Georg",
"middle": [],
"last": "Ziebertz",
"suffix": ""
}
],
"year": 2006,
"venue": "Social Behaviour and Personality",
"volume": "34",
"issue": "2",
"pages": "197--204",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Leslie Francis, Christopher Lewis, and Hans-Georg Ziebertz. 2006. The short-form revised Eysenenck personality questionnaire (EPQR-S): A German edi- tion. Social Behaviour and Personality, 34(2):197- 204.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Experiments with a new boosting algorithm",
"authors": [
{
"first": "Yoav",
"middle": [],
"last": "Freund",
"suffix": ""
},
{
"first": "Robert",
"middle": [],
"last": "Schapire",
"suffix": ""
}
],
"year": 1996,
"venue": "Thirteenth International Conference on Machine Learning",
"volume": "",
"issue": "",
"pages": "148--156",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yoav Freund and Robert Schapire. 1996. Experiments with a new boosting algorithm. In Thirteenth Interna- tional Conference on Machine Learning, pages 148- 156, San Francisco.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Who wrote the Bible?",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Friedmann",
"suffix": ""
}
],
"year": 1997,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Richard Friedmann. 1997. Who wrote the Bible? Harper, San Francisco.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Deriving marketing intelligence from online discussion",
"authors": [
{
"first": "Natalie",
"middle": [],
"last": "Glance",
"suffix": ""
},
{
"first": "Matthew",
"middle": [],
"last": "Hurst",
"suffix": ""
},
{
"first": "Kamal",
"middle": [],
"last": "Nigam",
"suffix": ""
},
{
"first": "Mathew",
"middle": [],
"last": "Siegler",
"suffix": ""
},
{
"first": "Robert",
"middle": [],
"last": "Stockton",
"suffix": ""
},
{
"first": "Takashi",
"middle": [],
"last": "Tomokiyo",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining",
"volume": "",
"issue": "",
"pages": "419--428",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Natalie Glance, Matthew Hurst, Kamal Nigam, Mathew Siegler, Robert Stockton, and Takashi Tomokiyo. 2005. Deriving marketing intelligence from online discussion. In Proceedings of the eleventh ACM SIGKDD international conference on Knowledge dis- covery in data mining, pages 419-428, Chicago.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Assessment of a significant Arabic corpus",
"authors": [
{
"first": "Abduelbaset",
"middle": [],
"last": "Goweder",
"suffix": ""
},
{
"first": "Anne",
"middle": [],
"last": "De Roeck",
"suffix": ""
}
],
"year": 2001,
"venue": "Proceedings of the ACL/EACL Workshop on Arabic Language Processing: Status and Prospects",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Abduelbaset Goweder and Anne De Roeck. 2001. As- sessment of a significant Arabic corpus. In Proceed- ings of the ACL/EACL Workshop on Arabic Language Processing: Status and Prospects, Toulouse.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Improvements to Platt's SMO algorithm for SVM classifier design",
"authors": [
{
"first": "Sathiya",
"middle": [],
"last": "Keerthi",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Shirish Shevade",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Bhattacharyya",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Murthy",
"suffix": ""
}
],
"year": 2001,
"venue": "Neural Computation",
"volume": "13",
"issue": "3",
"pages": "637--649",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sathiya Keerthi, Shirish Shevade, C. Bhattacharyya, and K. Murthy. 2001. Improvements to Platt's SMO algo- rithm for SVM classifier design. Neural Computation, 13(3):637-649.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Automatically determining an anonymous author's native language",
"authors": [
{
"first": "Moshe",
"middle": [],
"last": "Koppel",
"suffix": ""
},
{
"first": "Jonathan",
"middle": [],
"last": "Schler",
"suffix": ""
},
{
"first": "Kfir",
"middle": [],
"last": "Zigdon",
"suffix": ""
}
],
"year": 2005,
"venue": "editor, Intelligence and Security Informatics, Proceedings of the IEEE International Conference on Intelligence and Security Informatics (ISI 2005)",
"volume": "",
"issue": "",
"pages": "209--217",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Moshe Koppel, Jonathan Schler, and Kfir Zigdon. 2005. Automatically determining an anonymous author's na- tive language. In Paul B. Kantor et al., editor, In- telligence and Security Informatics, Proceedings of the IEEE International Conference on Intelligence and Security Informatics (ISI 2005), pages 209-217. Springer.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Shakespeare, Fletcher, and The Two Noble Kinsmen. Literary and Linguistic Computing",
"authors": [
{
"first": "Gerrard",
"middle": [],
"last": "Ledger",
"suffix": ""
},
{
"first": "Thomas",
"middle": [],
"last": "Merriam",
"suffix": ""
}
],
"year": 1994,
"venue": "",
"volume": "9",
"issue": "",
"pages": "235--248",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gerrard Ledger and Thomas Merriam. 1994. Shake- speare, Fletcher, and The Two Noble Kinsmen. Lit- erary and Linguistic Computing, 9(3):235-248.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Inference and Disputed Authorship: The Federalist. Series in behavioral science: Quantitative methods edition",
"authors": [
{
"first": "F",
"middle": [],
"last": "Mosteller",
"suffix": ""
},
{
"first": "D",
"middle": [
"L"
],
"last": "Wallace",
"suffix": ""
}
],
"year": 1964,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "F. Mosteller and D.L. Wallace. 1964. Inference and Dis- puted Authorship: The Federalist. Series in behav- ioral science: Quantitative methods edition. Addison- Wesley.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Toward an adequate taxonomy of personality attributes: replicated factor structure in peer nomination personality rating",
"authors": [
{
"first": "T",
"middle": [],
"last": "Warren",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Norman",
"suffix": ""
}
],
"year": 1963,
"venue": "Journal of Abnormal and Social Psychology",
"volume": "66",
"issue": "",
"pages": "574--583",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Warren T. Norman. 1963. Toward an adequate taxonomy of personality attributes: replicated factor structure in peer nomination personality rating. Journal of Abnor- mal and Social Psychology, 66:574-583.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "Whose thumb is it anyway? Classifying author personality from weblog text",
"authors": [
{
"first": "Jon",
"middle": [],
"last": "Oberlander",
"suffix": ""
},
{
"first": "Scott",
"middle": [],
"last": "Nowson",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the COLING/ACL 2006 Main Conference Poster Sessions",
"volume": "",
"issue": "",
"pages": "627--634",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jon Oberlander and Scott Nowson. 2006. Whose thumb is it anyway? Classifying author personality from weblog text. In Proceedings of the COLING/ACL 2006 Main Conference Poster Sessions, pages 627- 634, Sydney.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "C4.5: Programs for Machine Learning",
"authors": [
{
"first": "Ross",
"middle": [],
"last": "Quinlan",
"suffix": ""
}
],
"year": 1993,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ross Quinlan. 1993. C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo.",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "Data Mining: Practical machine learning tools and techniques",
"authors": [
{
"first": "H",
"middle": [],
"last": "Ian",
"suffix": ""
},
{
"first": "Eibe",
"middle": [],
"last": "Witten",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Frank",
"suffix": ""
}
],
"year": 2005,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ian H. Witten and Eibe Frank. 2005. Data Mining: Prac- tical machine learning tools and techniques. Morgan Kaufmann, San Francisco, 2nd edition.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "Authorship analysis in cybercrime investigation",
"authors": [
{
"first": "Rong",
"middle": [],
"last": "Zheng",
"suffix": ""
},
{
"first": "Yi",
"middle": [],
"last": "Qin",
"suffix": ""
},
{
"first": "Zan",
"middle": [],
"last": "Huang",
"suffix": ""
},
{
"first": "Hsinchun",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the first NSF/NIJ Intelligence and Security Informatics Symposium",
"volume": "",
"issue": "",
"pages": "59--73",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rong Zheng, Yi Qin, Zan Huang, and Hsinchun Chen. 2003. Authorship analysis in cybercrime investiga- tion. In Hsinchun Chen et al., editor, Proceedings of the first NSF/NIJ Intelligence and Security Informatics Symposium, pages 59-73. Springer.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "A framework for authorship identification of online messages: Writing-style features and classification techniques",
"authors": [
{
"first": "Rong",
"middle": [],
"last": "Zheng",
"suffix": ""
},
{
"first": "Jiexun",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Hsinchun",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Zan",
"middle": [],
"last": "Huang",
"suffix": ""
}
],
"year": 2006,
"venue": "JASIST",
"volume": "57",
"issue": "3",
"pages": "378--393",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rong Zheng, Jiexun Li, Hsinchun Chen, and Zan Huang. 2006. A framework for authorship identification of online messages: Writing-style features and classifi- cation techniques. JASIST, 57(3):378-393.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"uris": null,
"text": "TAT System Diagram tations and saved in a common structure called the Annotation Repository.",
"num": null
},
"TABREF1": {
"content": "<table/>",
"type_str": "table",
"num": null,
"text": "Overview of the collected English and Arabic email data.",
"html": null
},
"TABREF3": {
"content": "<table><tr><td>all</td><td>81.15</td><td>0.00</td></tr><tr><td colspan=\"2\">all except arabicNamedEntities 80.79</td><td>-0.36</td></tr><tr><td colspan=\"2\">all except arabicMorphological 80.19</td><td>-0.96</td></tr><tr><td>all except arabicChar</td><td>79.99</td><td>-1.16</td></tr><tr><td>all except arabicLexical</td><td>77.44</td><td>-3.71</td></tr></table>",
"type_str": "table",
"num": null,
"text": "",
"html": null
},
"TABREF4": {
"content": "<table><tr><td>Trait</td><td colspan=\"2\">English email data</td><td colspan=\"2\">Arabic email data</td></tr><tr><td/><td colspan=\"4\">Accuracy Improvement over Baseline Accuracy Improvement over Baseline</td></tr><tr><td>Age</td><td>56.46</td><td>+17.03</td><td>72.10</td><td>+02.01</td></tr><tr><td>Gender</td><td>69.26</td><td>+14.78</td><td>81.15</td><td>+08.99</td></tr><tr><td colspan=\"2\">Education 79.92</td><td>+21.14</td><td>93.66</td><td>+00.04</td></tr></table>",
"type_str": "table",
"num": null,
"text": "Contribution of feature groups to improvements in gender prediction.",
"html": null
},
"TABREF5": {
"content": "<table/>",
"type_str": "table",
"num": null,
"text": "Comparison of results for demographic traits for English and Arabic.",
"html": null
}
}
}
} |