File size: 110,679 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 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 |
{
"paper_id": "I17-1032",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T07:38:08.983083Z"
},
"title": "ES-LDA: Entity Summarization using Knowledge-based Topic Modeling",
"authors": [
{
"first": "Seyedamin",
"middle": [],
"last": "Pouriyeh",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Georgia",
"location": {
"settlement": "Athens",
"region": "GA",
"country": "USA"
}
},
"email": "pouriyeh@uga.edu"
},
{
"first": "Mehdi",
"middle": [],
"last": "Allahyari",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Sothern University",
"location": {
"settlement": "Statesboro",
"country": "Georgia, USA"
}
},
"email": "mallahyari@georgiasouthern.edu"
},
{
"first": "Krys",
"middle": [],
"last": "Kochut",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Georgia",
"location": {
"settlement": "Athens",
"region": "GA",
"country": "USA"
}
},
"email": "kkochut@uga.edu"
},
{
"first": "",
"middle": [],
"last": "Gong Cheng",
"suffix": "",
"affiliation": {
"laboratory": "National Key Laboratory for Novel Software Technology",
"institution": "Nanjing University",
"location": {
"settlement": "Nanjing",
"country": "China"
}
},
"email": ""
},
{
"first": "Hamid",
"middle": [
"Reza"
],
"last": "Arabnia",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Georgia",
"location": {
"settlement": "Athens",
"region": "GA",
"country": "USA"
}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "With the advent of the Internet, the amount of Semantic Web documents that describe real-world entities and their inter-links as a set of statements have grown considerably. These descriptions are usually lengthy, which makes the utilization of the underlying entities a difficult task. Entity summarization, which aims to create summaries for real world entities, has gained increasing attention in recent years. In this paper, we propose a probabilistic topic model, ES-LDA, that combines prior knowledge with statistical learning techniques within a single framework to create more reliable and representative summaries for entities. We demonstrate the effectiveness of our approach by conducting extensive experiments and show that our model outperforms the state-of-the-art techniques and enhances the quality of the entity summaries.",
"pdf_parse": {
"paper_id": "I17-1032",
"_pdf_hash": "",
"abstract": [
{
"text": "With the advent of the Internet, the amount of Semantic Web documents that describe real-world entities and their inter-links as a set of statements have grown considerably. These descriptions are usually lengthy, which makes the utilization of the underlying entities a difficult task. Entity summarization, which aims to create summaries for real world entities, has gained increasing attention in recent years. In this paper, we propose a probabilistic topic model, ES-LDA, that combines prior knowledge with statistical learning techniques within a single framework to create more reliable and representative summaries for entities. We demonstrate the effectiveness of our approach by conducting extensive experiments and show that our model outperforms the state-of-the-art techniques and enhances the quality of the entity summaries.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "With the emergence of Linked Open Data (LOD) 1 as a way of publishing and interacting with the information, many datasets such as DBpedia (Bizer et al., 2009) and YAGO (Hoffart et al., 2013) have been created and are publicly available on the Web. For example, DBpedia as part of LOD is a knowledge base extracted from Wikipedia that consists of Wikipedia resources (entities) described as RDF statements (i.e., RDF triples). The Resource Description Framework (RDF) is the Semantic Web standard data model used for representing information on the Web. An RDF triple is represented in the form of < subject, predicate, object >. The latest English version of DBpedia contains over 4.5 million entities collectively described by over 1.6 billion triples. This means that each entity description has an average of 355 RDF triples. Human users and computer applications need to consider these lengthy descriptions while performing various semantic tasks. Thus, entity summarization, a task of producing more concise, but still sufficient entity description, has garnered a significant amount of attention.",
"cite_spans": [
{
"start": 138,
"end": 158,
"text": "(Bizer et al., 2009)",
"ref_id": "BIBREF4"
},
{
"start": 168,
"end": 190,
"text": "(Hoffart et al., 2013)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Recently, with the huge growth of information, summarization techniques are becoming some of the main approaches to making the information more readily available. In fact, summarization techniques aim to facilitate the identification of structure and meaning in data. Researchers in different communities have taken a strong interest in this task and, accordingly, have proposed various methods for a wide variety of summarization techniques in multiple areas. Document summarization (Nenkova and McKeown, 2012) , database summarization (Bu et al., 2005) , and graph summarization (Navlakha et al., 2008) are just a few examples that have been studied by different communities. RDF data summarization and in particular entity summarization, has attracted considerable attentions in recent years as it can benefit many other tasks in the natural language processing area, including entity recognition (Zhao and Kit, 2008) , entity disambiguation (Dai et al., 2011) , and many others. Several approaches have been developed to summarize RDF data with respect to entities, including RELIN (Cheng et al., 2011) , FACES (Gunaratna et al., 2015) , and LinkSUM (Thalhammer et al., 2016) . RDF summarization differs from document summarization in the sense that RDF triples are structured and do not have many frequently used words to help the summarization task, which makes RDF summarization more challenging.",
"cite_spans": [
{
"start": 484,
"end": 511,
"text": "(Nenkova and McKeown, 2012)",
"ref_id": "BIBREF19"
},
{
"start": 537,
"end": 554,
"text": "(Bu et al., 2005)",
"ref_id": "BIBREF7"
},
{
"start": 581,
"end": 604,
"text": "(Navlakha et al., 2008)",
"ref_id": "BIBREF18"
},
{
"start": 900,
"end": 920,
"text": "(Zhao and Kit, 2008)",
"ref_id": "BIBREF29"
},
{
"start": 945,
"end": 963,
"text": "(Dai et al., 2011)",
"ref_id": "BIBREF9"
},
{
"start": 1086,
"end": 1106,
"text": "(Cheng et al., 2011)",
"ref_id": "BIBREF8"
},
{
"start": 1115,
"end": 1139,
"text": "(Gunaratna et al., 2015)",
"ref_id": "BIBREF14"
},
{
"start": 1154,
"end": 1179,
"text": "(Thalhammer et al., 2016)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Topic modeling has become a popular method for uncovering the hidden themes from text corpora. Topic models usually consider each document as a mixture of topics, where a topic is a probability distribution over words. When the topic proportions of documents are estimated, they can be used as the themes (high-level semantics) of the documents. Topic models have been widely used for various text mining tasks, such as machine translation (Su et al., 2015) , word embedding (Batmanghelich et al., 2016; Das et al., 2015) , automatic topic labeling (Wan and Wang, 2016; Allahyari and Kochut, 2015; Allahyari et al., 2017b) , and others (Allahyari et al., 2017a) .",
"cite_spans": [
{
"start": 440,
"end": 457,
"text": "(Su et al., 2015)",
"ref_id": "BIBREF22"
},
{
"start": 475,
"end": 503,
"text": "(Batmanghelich et al., 2016;",
"ref_id": "BIBREF3"
},
{
"start": 504,
"end": 521,
"text": "Das et al., 2015)",
"ref_id": "BIBREF10"
},
{
"start": 549,
"end": 569,
"text": "(Wan and Wang, 2016;",
"ref_id": "BIBREF26"
},
{
"start": 570,
"end": 597,
"text": "Allahyari and Kochut, 2015;",
"ref_id": "BIBREF0"
},
{
"start": 598,
"end": 622,
"text": "Allahyari et al., 2017b)",
"ref_id": "BIBREF2"
},
{
"start": 636,
"end": 661,
"text": "(Allahyari et al., 2017a)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this paper, we propose a novel topic model, called ES-LDA, that integrates prior knowledge with the topic modeling within a single framework for RDF entity summarization. In our approach, each entity, which is considered as a document, is a multinomial distribution over the predicates (properties), where each predicate is a probability distribution over the subjects and objects of the triples in the RDF data. We rank the triples based on their probability distributions and choose the top-k triples that best describe the underlying entity as its summary. We evaluated our approach against state-of-the-art techniques and our experiments indicate that our approach outperforms other methods in terms of the quality of summarization.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The rest of the paper is organized as follows: Section 2 presents an overview of related work. Section 3 introduces the baseline for this paper. In Section 4, we define the main problem and propose our model in detail and afterwards, in Section 5, we explain the configurations of our model and describe the experiments. Finally, in Sections 6 and Section 7, we discuss the results and conclude the paper, respectively.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Summarization methods can be divided into two main categories, which are called extractive and none-extractive (abstractive) summarization. In extractive approaches, which are usually applicable in text and ontology summarization (Jones, 2007) (Zhang et al., 2007) , a set of features is extracted directly from the input data. On the other hand, in non-extractive methods, which generally are employed in graph (Navlakha et al., 2008) and database (Bu et al., 2005) summarization, new sentences from the input data are generated (Hahn and Mani, 2000) to form a summary. In this research, we focus on extractive summarization. The concept of entity summarization in the form of RDF graph data has attracted more attention in recent years. Cheng et al. (Cheng et al., 2011) proposed entity summarization method, called RE-LIN, based on the PageRank algorithm to extract representative triples, called representative features for RDF graph entities. Because of the centrality based ranking issue, RELIN highlights the most similar and central triples, while in summarization, the diversity of summarized triples is the key point. SUMMARUM (Thalhammer and Rettinger, 2014) is a system for a better navigation within Linked Data through the ranking of triples. This system also uses the PageRank algorithm to rank triples according to the popularity of resources with the help of Wikipedia pages. Two aforementioned approaches could not meet the diversity requirement in the summarization process. FACES (Gunaratna et al., 2015) , on the other hand, tries to keep a balance between the centrality and diversity of the selected triples for each entity. It utilizes a clustering algorithm, called Cobweb (Fisher, 1987) , to cluster related triples before ranking them to keep the diversity in the summarization. The recent version of SUMMARUM, which is called LinkSUM (Thalhammer et al., 2016) , focused more on the objects instead of the diversity of properties for entities and showed a better result on the same dataset, in comparison with FACES. Beside the aforementioned techniques dedicated to entity summarization, there are various ranking models and tools, including TripleRank (Franz et al., 2009) and TRank (Tonon et al., 2013) that rank triples and concepts, respectively, incorporating ranking algorithms. However, Cheng et al. (2011) indicated that these methods are not appropriate for the entity summarization problem, which needs ranking of feature sets based on their importance to identify the underlying entity.",
"cite_spans": [
{
"start": 230,
"end": 243,
"text": "(Jones, 2007)",
"ref_id": "BIBREF17"
},
{
"start": 244,
"end": 264,
"text": "(Zhang et al., 2007)",
"ref_id": "BIBREF28"
},
{
"start": 412,
"end": 435,
"text": "(Navlakha et al., 2008)",
"ref_id": "BIBREF18"
},
{
"start": 449,
"end": 466,
"text": "(Bu et al., 2005)",
"ref_id": "BIBREF7"
},
{
"start": 530,
"end": 551,
"text": "(Hahn and Mani, 2000)",
"ref_id": "BIBREF15"
},
{
"start": 739,
"end": 772,
"text": "Cheng et al. (Cheng et al., 2011)",
"ref_id": "BIBREF8"
},
{
"start": 1137,
"end": 1169,
"text": "(Thalhammer and Rettinger, 2014)",
"ref_id": "BIBREF24"
},
{
"start": 1500,
"end": 1524,
"text": "(Gunaratna et al., 2015)",
"ref_id": "BIBREF14"
},
{
"start": 1698,
"end": 1712,
"text": "(Fisher, 1987)",
"ref_id": "BIBREF11"
},
{
"start": 1862,
"end": 1887,
"text": "(Thalhammer et al., 2016)",
"ref_id": "BIBREF23"
},
{
"start": 2181,
"end": 2201,
"text": "(Franz et al., 2009)",
"ref_id": "BIBREF12"
},
{
"start": 2212,
"end": 2232,
"text": "(Tonon et al., 2013)",
"ref_id": "BIBREF25"
},
{
"start": 2322,
"end": 2341,
"text": "Cheng et al. (2011)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "An RDF data graph is a collection of nodes and edges that connect the nodes together. Nodes are usually recognized by unique IDs which are called Uniform Resource Identifiers (URIs) or exact values (i.e. numbers, dates, etc) namely Lit- Table 1 : J.C.Penny entity predicates and corresponding objects with the top-5 ES-LDA summary.",
"cite_spans": [],
"ref_spans": [
{
"start": 237,
"end": 244,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Preliminaries",
"sec_num": "3"
},
{
"text": "erals. An RDF graph is represented in a form of a collection of triples, each including a Subject, Predicate, and Object. In an RDF graph, an entity is defined as a subject with all predicates and corresponding objects to those predicates, collectively forming the entity's description. As Table 1 shows, the J.C.Penny entity is represented by its predicates (properties) and the corresponding objects in the triple format. For example, the triple < J.C.P enny, industry, Retail > introduces J.C.Penny's industry as Retail (due to space limitations we have dropped the first part of the URIs).",
"cite_spans": [],
"ref_spans": [
{
"start": 290,
"end": 297,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Preliminaries",
"sec_num": "3"
},
{
"text": "Definition 1 (Entity summary): Given an entity e and a positive integer k, a summary of the entity e, denoted Sum(e, k), is the top-k subset of all predicates and corresponding objects that are most relevant to that entity. As Table1 shows the top-5 summary for J.C.Penny entity, which is represented through foundedBy, industry, keyPerson, location, and type.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Preliminaries",
"sec_num": "3"
},
{
"text": "The Latent Dirichlet Allocation (LDA) is a generative probabilistic model for extracting thematic information (topics) from a collection of documents. LDA assumes that each document is made up of various topics, where each topic is a probability distribution over words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "Let",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "D = {d 1 , d 2 , . . . , d |D| } be a corpus of documents and V = {w 1 , w 2 , . . . , w |V| } a vocabulary (words) of the corpus. A topic z j , 1 \u2264 j \u2264 K is represented as a multino- mial probability distribution over the |V| words, p(w i |z j ), |V| i p(w i |z j ) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "LDA generates the words in a two-stage process: words are gener-",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "\u21b5 \u2713 z w N D K 1 Figure 1: LDA Graphical Representation",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "ated from topics and topics are generated by documents. More formally, the distribution of words, given the document, is calculated as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "p(w i |d) = K j=1 p(w i |z j )p(z j |d)",
"eq_num": "(1)"
}
],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "The graphical model of LDA is shown in Figure 1 and the generative process for the corpus D is:",
"cite_spans": [],
"ref_spans": [
{
"start": 39,
"end": 48,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "1. For each topic k \u2208 {1, 2, . . . , K}, sample a word distribution \u03c6 k \u223c Dir(\u03b2) 2. For each document d \u2208 {1, 2, . . . , D}, (a) Sample a topic distribution \u03b8 d \u223c Dir(\u03b1) (b) For each word w n , where n \u2208 {1, 2, . . . , N }, in document d, i. Sample a topic z i \u223c Mult(\u03b8 d ) ii. Sample a word w n \u223c Mult(\u03c6 z i )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "In the LDA model, the word-topic distribution p(w|z) and topic-document distribution p(z|d) are learned entirely in an unsupervised manner, without any prior knowledge about what words are re-lated to the topics and what topics are related to individual documents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Latent Dirichlet Allocation (LDA)",
"sec_num": "3.1"
},
{
"text": "In this section, we first describe the problem and then define how to utilize topic models for RDF graphs. Then, we formally introduce our ES-LDA model and explain how to integrate prior knowledge from RDF data graph within a topic model for entity summarization.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Problem Statement",
"sec_num": "4"
},
{
"text": "Generating summaries for voluminous Semantic Web data, and in particular RDF data, for quick identification of entities has gained considerable attention as a challenging problem in the Semantic Web community. In the literature, Entity Summarization is defined as selecting a small but representative subset of the original triples associated with an entity. In this context, given an RDF data set comprising a collection of entities, where each entity is described by a set of its properties (i.e., all triples with the entity as the subject), our goal is to choose top-k representative triples for each entity. In other words, since all triples associated with an entity (as its description) share the same subject, our objective is to select top-k predicates and their corresponding objects among these triples that best summarize the entity's description.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Problem Definition",
"sec_num": "4.1"
},
{
"text": "Topic models were originally introduced for text documents, however, they have been applied to other types of data, such as images , and recently (Sleeman et al., 2015) used topic modeling for RDF graphs. The first step in applying topic models is to define documents and word-like elements as the basic building blocks of documents. Since an RDF graph is usually represented as a set of triples, where each triple t consists of a subject s, predicate p, and an object o, in the form of <s, p, o>, we can consider a collection of such triples as a \"document\".",
"cite_spans": [
{
"start": 146,
"end": 168,
"text": "(Sleeman et al., 2015)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Topic Models for RDF Graphs",
"sec_num": "4.2"
},
{
"text": "Definition 2 (document): A document d is defined as a set of triples,",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Topic Models for RDF Graphs",
"sec_num": "4.2"
},
{
"text": "d = {t 1 , t 2 , \u2022 \u2022 \u2022 , t n },",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Topic Models for RDF Graphs",
"sec_num": "4.2"
},
{
"text": "that describe a single entity e. In other words, all triples of a document d have the same subject.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Topic Models for RDF Graphs",
"sec_num": "4.2"
},
{
"text": "\"Words\" of a document can be extracted from different parts of its triples. We define a \"word\" w as the subject or object of a triple t in document d. Therefore, each document is represented by a \"bag of words\" including all the subjects and objects of its triples. In this paper, all subjects in the triples of a document are the same, because each document corresponds to a single entity, hence, in practice each document is a \"bag of objects\" 2 Topic models usually utilize some data preprocessing, such as punctuation removal, downcasting, and abbreviation expansion, etc., to enhance the final performance. We also performed preprocessing on the RDF data and filtered out the schema and dataset dependent predicates, such as sameAs, wikiPageExternalLink, subject, wikiPageWikiLink, in addition to literals. Since we work with RDF graphs that differ from typical text documents in the sense that RDF data are represented as triples, we need to address several challenges mentioned in (Sleeman et al., 2015) to be able to run topic models on RDF data. These challenges include sparseness, use of unnatural language, and the lack of context. RDF data can be affected by Sparseness. We consider documents as sets of triples associated with a single entity. Such a set can be very large, leading to a large bag of words with a semantic theme, or small (sparse), resulting in a poor bag of words with less contextual information. It is also possible that a document with a high number of triples ends up having a small bag of words after pre-processing; for example based on Table 1 , J.C.Penny entity comes with United States, James Cash Penney, Retail, Ron Johnson, Plano, Texas, United States, S&P 500 and Public company as a bag of words for J.C.Penny entity, which shows sparseness in this document. Unnatural Language can be problematic for RDF data. A typical text document contains sentences where each sentence has a natural structure. These extra components of a sentence usually provide a further \"context\" for understanding words that are ambiguous or have multiple meanings, such as polysemous or homonymous ones. The aforementioned example for the J.C.Penny entity also confirms the unnatural language problem. The \"lack of context\" can further impact RDF data because they are potentially sparse, described by unnatural language, and often using words that have multiple meanings, difficult to differentiate (J.C.Penny bag of words example). Additionally, triples are more prone to pre-processing, because it is not uncom- RDF data resemble short texts in terms of the aforementioned challenges. Sparseness in a short text causes the model to be less discriminative to recognize how words are related and the limited context makes it hard for the model to identify the meanings of the words in such short text documents (Yan et al., 2013) . In order to alleviate these issues, researchers usually take two approaches. They either augment the short text or design custom versions of the LDA model that address their specific problems. In this paper, we have used both approaches. We describe how to supplement the RDF data in the following section and describe the details of our model in section 4.4.",
"cite_spans": [
{
"start": 988,
"end": 1010,
"text": "(Sleeman et al., 2015)",
"ref_id": "BIBREF21"
},
{
"start": 2834,
"end": 2852,
"text": "(Yan et al., 2013)",
"ref_id": "BIBREF27"
}
],
"ref_spans": [
{
"start": 1574,
"end": 1581,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Topic Models for RDF Graphs",
"sec_num": "4.2"
},
{
"text": "As topic modeling is based on statistics of the cooccurrence of terms (Sleeman et al., 2015) , when we are dealing with short texts with a very limited number of repetitions, which is the case with RDF data, we need to find a way to supplement the data to elevate the performance of the topic modeling approach. We augment the documents using two different methods. In the first method, we increase the frequency of the words in each document. But the question is \"How many times each word of a document should be repeated?\". Entities in DBpedia have been organized into a category network, therefore, every entity has a number of categories associated with it. The relationship between an entity and a category is defined by the \"http://purl.org/dc/terms/subject\" predicate. Since each word of a document is an object of a triple, and accordingly, an entity in DBpedia, it is related to several categories. We assume that objects (words) of a document that have more categories are likely more important. Thus, We expand each document by increasing the frequency of each object by the number of its categories. In the second method, instead of repeating each object a certain number of times, we enlarge each document by adding categories of the objects as extra words, directly to the document. There are multiple advantages of supplementing each document by adding object categories: (i) the sparseness in the document, related to each entity, is lowered as we are adding a number of related words to it; (ii) we reduce the ambiguity in the document, because adding extra categories alleviates the lack of context and helps distinguish the appropriate meanings of the words with multiple connotations; and lastly (iii), adding object categories makes the documents semantically more relevant to their topical themes. We evaluated our model using both methods and the results demonstrate that the first method gives significantly better summaries than the second method.",
"cite_spans": [
{
"start": 70,
"end": 92,
"text": "(Sleeman et al., 2015)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Supplementing RDF Data",
"sec_num": "4.3"
},
{
"text": "ES-LDA is a probabilistic generative model for modeling entities in RDF graphs. The key idea behind our model is twofold: (1) we exploit statistical topic models as the underlying quantitative framework for entity summarization; and (2) ES-LDA incorporates the prior knowledge from the RDF knowledge base directly into the topic model. The plate notation is shown in Figure 2 . In our model, each document is a multinomial distribution over the predicates. If we consider predicates as topics, at the document level, our model is the same as standard LDA. However, we set the number of topics in ES-LDA to be the number of unique predicates in the corpus. Unlike the standard LDA, where each topic is a multinomial distribution over the vocabulary from the Dirichlet prior \u03b2, in our model each predicate is a multinomial distribution over all the subjects and objects of the RDF graph. In our approach, a document consists of a set of triples describing a single entity, i.e. all these triples share the same subject. Thus, we constrain the documents to only have the objects of related triples and also restrict the predicates to be defined only over the objects. In addition, for each predicate r, we further smooth its distribution by \u039b r . \u039b is a matrix that has encoded the background knowledge about predicate-object values from DBpedia. Section 4.5 explains how \u039b is constructed. The generative process of ES-LDA is shown in Algorithm 1.",
"cite_spans": [],
"ref_spans": [
{
"start": 367,
"end": 375,
"text": "Figure 2",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Proposed Model",
"sec_num": "4.4"
},
{
"text": "Following this process, the joint probability of generating a corpus D = {d 1 , d 2 , . . . , d |D| }, the predicate assignments r given the hyperparameters \u03b1, \u03b2 and the prior matrix \u039b is:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Model",
"sec_num": "4.4"
},
{
"text": "P (o, s, r|\u03b1, \u03b2, \u039b) = \u03c6 P (\u03c6|\u03b2; \u039b) d r d P (o d |r d , \u03c6)P (s d |r d , \u03c6) \u00d7 \u03b8 P (\u03b8|\u03b1)P (r d |\u03b8, \u03c6)d\u03b8d\u03c6",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Model",
"sec_num": "4.4"
},
{
"text": "(2)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Model",
"sec_num": "4.4"
},
{
"text": "In the ES-LDA model, each predicate has a probability distribution over the objects of the RDF graph. Entity summarization is the task of choosing the top-k predicate-object pairs that best describe an entity. Presumably, if an object is associated with more categories in DBpedia, it is likely more important. We create the the \u039b matrix to encode the prior weight of the predicate-object pairs and utilize it to smooth the predicate-object distributions \u03c6 by incorporating this domain knowledge into the topic model. We build the \u039b matrix of size R \u00d7O, where R is the number of predicates and O is the number of objects in the RDF graph. Let f be an indicator function where f (i, j) = 1 if there is a triple in RDF graph with predicate i and object j, and 1 otherwise, for 1 \u2264 i \u2264 R and 1 \u2264 j \u2264 O. Additionally, let c be the number of categories assigned to object j. Then, we define \u039b ij as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Constructing Predicate-Object Prior Matrix \u039b",
"sec_num": "4.5"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "\u039b ij = c if f (i, j) = 1 1 otherwise.",
"eq_num": "(3)"
}
],
"section": "Constructing Predicate-Object Prior Matrix \u039b",
"sec_num": "4.5"
},
{
"text": "For example, the \"Barack Obama\" entity has multiple predicate-object pairs in DBpedia, including \"profession-author\", \"profession-lawyer\" and \"profession-professor\" pairs. According to DBpedia, c author = 2, c lawyer = 4 and c prof essor = 2. It is reasonable to expect a higher probability for the \"profession-lawyer\" pair as it seems to be slightly more important than the other two pairs for \"Barack Obama\". As a result, \u039b prof ession\u2212lawyer = 4, which promotes \"profession-lawyer\" in Eq. 5.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Constructing Predicate-Object Prior Matrix \u039b",
"sec_num": "4.5"
},
{
"text": "Since the posterior inference of the LDA is intractable, we need to find an algorithm for estimating the posterior inference. A variety of algorithms have been used to estimate the parameters of topic models, such as variational EM and Gibbs sampling (Griffiths and Steyvers, 2004) . In this paper we use the collapsed Gibbs sampling procedure for our ES-LDA topic model. Collapsed Gibbs sampling (Griffiths and Steyvers, 2004) is a Markov Chain Monte Carlo (MCMC) (Robert and Casella, 2004) algorithm, which constructs a Markov chain over the latent variables in the model and converges to the posterior distribution, after a number of iterations. In our case, we aim to construct a Markov chain that converges to the posterior distribution over r conditioned on observed subjects s, objects o, hyperparameters \u03b1, \u03b2, and the prior matrix \u039b.",
"cite_spans": [
{
"start": 251,
"end": 281,
"text": "(Griffiths and Steyvers, 2004)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "In our modified version of the learning algorithm to infer p(o i |r j ) and p(r j |d), we (1) constrain the objects that are not paired with a predicate to have 0 probability, i.e. p(o i |r j ) = 0, if (r i , o j ) / \u2208 RDF graph, and (2) P (s|r j ) = 1, since all the triples of a document have the same subject s. We derive the posterior inference from Eq. 2 as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "P (r|o, s, \u03b1, \u03b2, \u039b) = P (r, o, s|\u03b1, \u03b2, \u039b) P (o|\u03b1, \u03b2, \u039b) \u221d P (r, o|\u03b1, \u03b2, \u039b) \u221d P (r)P (o|r)P (s|r) (4) P (r i = r|o i = o, r \u2212i , o \u2212i , \u03b1, \u03b2, \u039b) \u221d n (d) r,\u2212i + \u03b1 r r (n (d) r ,\u2212i + \u03b1 r ) \u00d7 n (r) o,\u2212i + \u039b ro \u03b2 o o (n (r) o ,\u2212i + \u039b ro \u03b2 o )",
"eq_num": "(5)"
}
],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "where",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "n (r)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "o is the number of times object o is assigned to predicate r. n",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "(d) r",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "denotes the number of times predicate r is associated with document d. The subscript \u2212i indicates that the contribution of the current object o i being sampled is removed from the counts. After Gibbs sampling, we can use the sampled predicate to estimate the probability of a predicate, given a document, \u03b8 dr and the probability of an object, given a predicate, \u03c6 ro :",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "\u03b8 dr = n (d) r + \u03b1 r r (n (d) r + \u03b1 r )",
"eq_num": "(6)"
}
],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "\u03c6 ro = n (r) o + \u039b ro \u03b2 o o (n (r) o + \u039b ro \u03b2 o )",
"eq_num": "(7)"
}
],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "5 Experiments",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "We evaluated our ES-LDA model against the stateof-the-art LinkSUM (Thalhammer et al., 2016) and FACES (Gunaratna et al., 2015) systems. Our goal was to show that the ES-LDA model produces results that are closer to human judgment, in comparison with the other approaches. We used the same dataset 3 that was used in the experiments conducted with FACES, as well as LinkSUM models. The dataset contained 50 entities randomly selected from DBpedia (English version 3.9) in domains including politician, actors, scientist, song, film, country, city, river, company, game, etc.. 15 people in the field of Semantic Web were selected as reviewers and each entity was evaluated by at least 7 reviewers to produce the top-5 and top-10 summaries. The average number of properties for each entity was 44.",
"cite_spans": [
{
"start": 66,
"end": 91,
"text": "(Thalhammer et al., 2016)",
"ref_id": "BIBREF23"
},
{
"start": 102,
"end": 126,
"text": "(Gunaratna et al., 2015)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "Based on the two types of RDF supplement methods we discussed in 4.3, we applied two different configurations for the proposed model. In the first experiment, ES-LDA @config-1, we configured the system to supplement each entity (document) by repeating each object based on the number of categories that the object has in the DBpedia knowledge base. For example, for the triple < J.C.P enney, industry, Retail > we repeated Retail object, 5 times in that document, as Retail has five different categories in DBpedia (i.e. \"Retailers, Retailing, French words and phrases, Merchandising, Marketing\" )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "In the second experiment, ES-LDA @config-2, we configured the system to supplement each entity (document) by adding the corresponding category(ies) of each object into the document. In this case, each entity is defined as a bag of words including objects and categories of each object. For example, for the aforementioned triple, in addition to the Retail we included \"Retailers, Retailing, French words and phrases, Merchandising, Marketing\" as the corresponding categories to the Retail object.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "For the other parameters, we assumed a symmetric Dirichlet prior and set \u03b2 = 0.01 and \u03b1 = 50/R, where R is the total number of unique predicates. We ran the Gibbs sampling algorithm for 1000 iterations and computed the posterior inference after the last sampling iteration. We selected the top-5 and top-10 most probable properties for each entity and calculate the quality of the summary for each entity through equation 8.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "Quality(Sum(e)) = 1 n n i=1",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "|Sum(e) \u2229 Sum I i (e)| (8)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "In our experiments, we used the quality of the summary proposed in (Cheng et al., 2011) , in which n ideal summaries Sum I i (e) generated by expert users for i = 1, ..., n and the summaries generated by the system Sum(e) were compared. The average of the overlap between an ideal summary and a summary generated by the system is denoted as the quality of the summary, which is 0 \u2264 Quality(Sum(e)) \u2264 k in the top-k settings.",
"cite_spans": [
{
"start": 67,
"end": 87,
"text": "(Cheng et al., 2011)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Inference using Gibbs Sampling",
"sec_num": "4.6"
},
{
"text": "The summary in our model is defined as sets of representative triples that can summarize each entity (sets of triples with the same subject) in a way close to a human-created summary. We decided to use the last part of a URI to compare the generated summaries with the expert summaries and produce ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiment Results",
"sec_num": "5.1"
},
{
"text": "Top-5 Top-10 ES-LDA @ config-1 1.20 3.50 ES-LDA @ config-2 1.10",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Model",
"sec_num": null
},
{
"text": "3.26 LinkSUM@ config-1 1.20",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Model",
"sec_num": null
},
{
"text": "3.15 LinkSUM@ config-2 1.20 3.20 FACES 0.93 2.92 the Summary Quality for each entity and average them. As (Thalhammer et al., 2016) reproduced the FACES overall Summary Quality based on this criteria and also applied it to their model, we decided to use their result as it was completely aligned with our summary definition.",
"cite_spans": [
{
"start": 106,
"end": 131,
"text": "(Thalhammer et al., 2016)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Model",
"sec_num": null
},
{
"text": "In Table 2 , we compare the quality of the results from LinkSUM, FACES, and ES-LDA with two distinct configurations (supplementing by object reputation and object categories). As Table 2 shows, the quality of our model outperforms the FACES approach, in both cases. The ES-LDA @ config-2 demonstrates a comparable result with the two configurations of LinkSUM, while ES-LDA @ config-1 outperforms LinkSUM. For some of the entities, the predicates that ES-LDA selected as top-5 most probable did not exist in the FACES dataset. It forced us to calculate the quality of summary for some of the entities with just 4 predicates instead of 5. We believe to be the only reason why top-5 Quality of Summary was lower than or equal to LinkSUM. Although, we had the same issue for the top-10 results, overall, ES-LDA shows a better performance in two configurations.",
"cite_spans": [],
"ref_spans": [
{
"start": 3,
"end": 10,
"text": "Table 2",
"ref_id": "TABREF1"
},
{
"start": 179,
"end": 186,
"text": "Table 2",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Model",
"sec_num": null
},
{
"text": "We evaluated our approach against the stateof-the-art summarization techniques, including LinkSUM and FACES. LinkSUM primarily focuses on the most relevant facts for each entity, while FACES tries to keep a balance between diversity and relevancy in entity summarization. There is usually a trade-off between diversity and relevancy of the selected predicates. Our ES-LDA model maintains both diversity and relevancy, while representing each entity through top-k predicates. As shown in Table 2 , our model outperforms the state-of-the-art approaches. Table 3 illustrates a sample of entities from the dataset along with their top-10 predicates, for all approaches. As Table 3 shows, the LinkSUM model is focusing more on the objects, while predicate repetition is permitted. For example, <Marie Curie, birthPlace, Warsaw>, <Marie Curie, birthPlace, Russian Empire>, and <Marie Curie, birthPlace, Congress Polandare> are representing Marie Curie's birth place. Although, they differ in terms of objects, it is arguable that referring to the same predicate with multiple objects that are more likely relevant reduces the chance of other important triples that could potentially appear in the summary. It should be noted that in the current ES-LDA configuration, we have not considered predicate repetition, thus, all the predicates of the triples appearing in the resultant summary are unique. FACES on the other hand, considers predicate diversity and tries to keep a balance between the diversity and relevancy but the overall quality of the FACES model is lower than LinkSUM and ES-LDA. In the FACES model, there are selected predicates which seems to be less informative in the sense to be top-10 representative for a particular entity. For example, which is referring to a png file, could be replaced with more descriptive one. Additionally, our proposed technique features several unique characteristics: (1) the ES-LDA is a knowledgebased probabilistic model that combines prior knowledge with statistical learning technique into a unified framework for entity summarization;",
"cite_spans": [],
"ref_spans": [
{
"start": 487,
"end": 494,
"text": "Table 2",
"ref_id": "TABREF1"
},
{
"start": 552,
"end": 559,
"text": "Table 3",
"ref_id": "TABREF2"
},
{
"start": 669,
"end": 676,
"text": "Table 3",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "(2) for each entity, it ranks all predicates based on their importance by computing marginal probabilities for the predicates. Table 4 illustrates the top-5 predicates for a sample of two entities; and finally (3), each predicate can be represented as a probability distribution over objects in the ES-LDA model, which allows us to describe the relations (predicates) of the RDF graph based on its nodes as shown in Table 5 .",
"cite_spans": [],
"ref_spans": [
{
"start": 127,
"end": 134,
"text": "Table 4",
"ref_id": "TABREF3"
},
{
"start": 416,
"end": 423,
"text": "Table 5",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "We have proposed a knowledge-based probabilistic topic model, called ES-LDA, based on the RDF entity representation for entity summarization. In our experiments, we have applied two different configurations: one based on object repetitions and the other based on adding object's categories, to alleviate common RDF data problems including sparseness, unnatural language, and lack of context. We conducted extensive experiments, which show the quality of the top-10 triples in both configurations outperforms the state-of-the-art techniques, LinkSUM and FACES, while for the top-5 quality we surpassed FACES and equaled the LinkSUM results. There are many interesting future research directions of this work. It would be interesting to investigate how this model and a much richer set of topic models that combine prior knowledge with statistical learning techniques could be used for various tasks in the Semantic Web domain, such as ontology summarization, ontology tagging, and finding similar ontologies.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "7"
},
{
"text": "\"bag of words\" and \"bag of objects\" are interchangeably used.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "http://wiki.knoesis.org/index.php/FACES",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "Gong Cheng was partially funded by the NSFC under Grant 61572247.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Automatic topic labeling using ontology-based topic models",
"authors": [
{
"first": "Mehdi",
"middle": [],
"last": "Allahyari",
"suffix": ""
},
{
"first": "Krys",
"middle": [],
"last": "Kochut",
"suffix": ""
}
],
"year": 2015,
"venue": "14th International Conference on Machine Learning and Applications (ICMLA)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mehdi Allahyari and Krys Kochut. 2015. Automatic topic labeling using ontology-based topic models. In 14th International Conference on Machine Learn- ing and Applications (ICMLA), 2015. IEEE.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "A brief survey of text mining: Classification, clustering and extraction techniques",
"authors": [
{
"first": "Mehdi",
"middle": [],
"last": "Allahyari",
"suffix": ""
},
{
"first": "Seyedamin",
"middle": [],
"last": "Pouriyeh",
"suffix": ""
},
{
"first": "Mehdi",
"middle": [],
"last": "Assefi",
"suffix": ""
},
{
"first": "Saied",
"middle": [],
"last": "Safaei",
"suffix": ""
},
{
"first": "Elizabeth",
"middle": [
"D"
],
"last": "Trippe",
"suffix": ""
},
{
"first": "Juan",
"middle": [
"B"
],
"last": "Gutierrez",
"suffix": ""
},
{
"first": "Krys",
"middle": [],
"last": "Kochut",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1707.02919"
]
},
"num": null,
"urls": [],
"raw_text": "Mehdi Allahyari, Seyedamin Pouriyeh, Mehdi Assefi, Saied Safaei, Elizabeth D Trippe, Juan B Gutierrez, and Krys Kochut. 2017a. A brief survey of text min- ing: Classification, clustering and extraction tech- niques. arXiv preprint arXiv:1707.02919 .",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "A knowledgebased topic modeling approach for automatic topic labeling",
"authors": [
{
"first": "Mehdi",
"middle": [],
"last": "Allahyari",
"suffix": ""
},
{
"first": "Seyedamin",
"middle": [],
"last": "Pouriyeh",
"suffix": ""
},
{
"first": "Krys",
"middle": [],
"last": "Kochut",
"suffix": ""
},
{
"first": "Hamid",
"middle": [
"R"
],
"last": "Arabnia",
"suffix": ""
}
],
"year": 2017,
"venue": "INTERNATIONAL JOURNAL OF AD-VANCED COMPUTER SCIENCE AND APPLICA-TIONS",
"volume": "8",
"issue": "9",
"pages": "335--349",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mehdi Allahyari, Seyedamin Pouriyeh, Krys Kochut, and Hamid R Arabnia. 2017b. A knowledge- based topic modeling approach for automatic topic labeling. INTERNATIONAL JOURNAL OF AD- VANCED COMPUTER SCIENCE AND APPLICA- TIONS 8(9):335-349.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Nonparametric spherical topic modeling with word embed",
"authors": [
{
"first": "Kayhan",
"middle": [],
"last": "Batmanghelich",
"suffix": ""
},
{
"first": "Ardavan",
"middle": [],
"last": "Saeedi",
"suffix": ""
},
{
"first": "Karthik",
"middle": [],
"last": "Narasimhan",
"suffix": ""
},
{
"first": "Sam",
"middle": [],
"last": "Gershman",
"suffix": ""
}
],
"year": 2016,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1604.00126"
]
},
"num": null,
"urls": [],
"raw_text": "Kayhan Batmanghelich, Ardavan Saeedi, Karthik Narasimhan, and Sam Gershman. 2016. Nonpara- metric spherical topic modeling with word embed- dings. arXiv preprint arXiv:1604.00126 .",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Dbpedia-a crystallization point for the web of data",
"authors": [
{
"first": "Christian",
"middle": [],
"last": "Bizer",
"suffix": ""
},
{
"first": "Jens",
"middle": [],
"last": "Lehmann",
"suffix": ""
},
{
"first": "Georgi",
"middle": [],
"last": "Kobilarov",
"suffix": ""
},
{
"first": "S\u00f6ren",
"middle": [],
"last": "Auer",
"suffix": ""
},
{
"first": "Christian",
"middle": [],
"last": "Becker",
"suffix": ""
},
{
"first": "Richard",
"middle": [],
"last": "Cyganiak",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "Hellmann",
"suffix": ""
}
],
"year": 2009,
"venue": "Web Semantics: science, services and agents on the world wide web",
"volume": "7",
"issue": "",
"pages": "154--165",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Christian Bizer, Jens Lehmann, Georgi Kobilarov, S\u00f6ren Auer, Christian Becker, Richard Cyganiak, and Sebastian Hellmann. 2009. Dbpedia-a crystal- lization point for the web of data. Web Semantics: science, services and agents on the world wide web 7(3):154-165.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Modeling annotated data",
"authors": [
{
"first": "M",
"middle": [],
"last": "David",
"suffix": ""
},
{
"first": "Michael I Jordan",
"middle": [],
"last": "Blei",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the 26th annual international ACM SIGIR conference on Research and development in informaion retrieval",
"volume": "",
"issue": "",
"pages": "127--134",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "David M Blei and Michael I Jordan. 2003. Model- ing annotated data. In Proceedings of the 26th an- nual international ACM SIGIR conference on Re- search and development in informaion retrieval. ACM, pages 127-134.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Mdl summarization with holes",
"authors": [
{
"first": "Shaofeng",
"middle": [],
"last": "Bu",
"suffix": ""
},
{
"first": "V",
"middle": [
"S"
],
"last": "Laks",
"suffix": ""
},
{
"first": "Raymond T",
"middle": [],
"last": "Lakshmanan",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ng",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the 31st international conference on Very large data bases. VLDB Endowment",
"volume": "",
"issue": "",
"pages": "433--444",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Shaofeng Bu, Laks VS Lakshmanan, and Raymond T Ng. 2005. Mdl summarization with holes. In Pro- ceedings of the 31st international conference on Very large data bases. VLDB Endowment, pages 433-444.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Relin: relatedness and informativeness-based centrality for entity summarization. The Semantic Web-ISWC",
"authors": [
{
"first": "Gong",
"middle": [],
"last": "Cheng",
"suffix": ""
},
{
"first": "Thanh",
"middle": [],
"last": "Tran",
"suffix": ""
},
{
"first": "Yuzhong",
"middle": [],
"last": "Qu",
"suffix": ""
}
],
"year": 2011,
"venue": "",
"volume": "",
"issue": "",
"pages": "114--129",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gong Cheng, Thanh Tran, and Yuzhong Qu. 2011. Re- lin: relatedness and informativeness-based centrality for entity summarization. The Semantic Web-ISWC 2011 pages 114-129.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Entity disambiguation using a markov-logic network",
"authors": [
{
"first": "Hong-Jie",
"middle": [],
"last": "Dai",
"suffix": ""
},
{
"first": "Richard",
"middle": [],
"last": "Tzong-Han",
"suffix": ""
},
{
"first": "Wen-Lian",
"middle": [],
"last": "Tsai",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Hsu",
"suffix": ""
}
],
"year": 2011,
"venue": "IJCNLP",
"volume": "",
"issue": "",
"pages": "846--855",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hong-Jie Dai, Richard Tzong-Han Tsai, Wen-Lian Hsu, et al. 2011. Entity disambiguation using a markov-logic network. In IJCNLP. pages 846-855.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Gaussian lda for topic models with word embeddings",
"authors": [
{
"first": "Rajarshi",
"middle": [],
"last": "Das",
"suffix": ""
},
{
"first": "Manzil",
"middle": [],
"last": "Zaheer",
"suffix": ""
},
{
"first": "Chris",
"middle": [],
"last": "Dyer",
"suffix": ""
}
],
"year": 2015,
"venue": "ACL (1)",
"volume": "",
"issue": "",
"pages": "795--804",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rajarshi Das, Manzil Zaheer, and Chris Dyer. 2015. Gaussian lda for topic models with word embed- dings. In ACL (1). pages 795-804.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Knowledge acquisition via incremental conceptual clustering",
"authors": [
{
"first": "H",
"middle": [],
"last": "Douglas",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Fisher",
"suffix": ""
}
],
"year": 1987,
"venue": "Machine learning",
"volume": "2",
"issue": "2",
"pages": "139--172",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Douglas H Fisher. 1987. Knowledge acquisition via in- cremental conceptual clustering. Machine learning 2(2):139-172.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Triplerank: Ranking semantic web data by tensor decomposition. The Semantic Web-ISWC",
"authors": [
{
"first": "Thomas",
"middle": [],
"last": "Franz",
"suffix": ""
},
{
"first": "Antje",
"middle": [],
"last": "Schultz",
"suffix": ""
},
{
"first": "Sergej",
"middle": [],
"last": "Sizov",
"suffix": ""
},
{
"first": "Steffen",
"middle": [],
"last": "Staab",
"suffix": ""
}
],
"year": 2009,
"venue": "",
"volume": "",
"issue": "",
"pages": "213--228",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Thomas Franz, Antje Schultz, Sergej Sizov, and Stef- fen Staab. 2009. Triplerank: Ranking semantic web data by tensor decomposition. The Semantic Web- ISWC 2009 pages 213-228.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Finding scientific topics",
"authors": [
{
"first": "L",
"middle": [],
"last": "Thomas",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Griffiths",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Steyvers",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the National academy of Sciences of the United States of America",
"volume": "101",
"issue": "1",
"pages": "5228--5235",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Thomas L Griffiths and Mark Steyvers. 2004. Find- ing scientific topics. Proceedings of the National academy of Sciences of the United States of America 101(Suppl 1):5228-5235.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Faces: diversity-aware entity summarization using incremental hierarchical conceptual clustering",
"authors": [
{
"first": "Kalpa",
"middle": [],
"last": "Gunaratna",
"suffix": ""
},
{
"first": "Krishnaprasad",
"middle": [],
"last": "Thirunarayan",
"suffix": ""
},
{
"first": "Amit P",
"middle": [],
"last": "Sheth",
"suffix": ""
}
],
"year": 2015,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kalpa Gunaratna, Krishnaprasad Thirunarayan, and Amit P Sheth. 2015. Faces: diversity-aware entity summarization using incremental hierarchical con- ceptual clustering .",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "The challenges of automatic summarization",
"authors": [
{
"first": "Udo",
"middle": [],
"last": "Hahn",
"suffix": ""
},
{
"first": "Inderjeet",
"middle": [],
"last": "Mani",
"suffix": ""
}
],
"year": 2000,
"venue": "Computer",
"volume": "33",
"issue": "11",
"pages": "29--36",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Udo Hahn and Inderjeet Mani. 2000. The challenges of automatic summarization. Computer 33(11):29-36.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Yago2: A spatially and temporally enhanced knowledge base from wikipedia",
"authors": [
{
"first": "Johannes",
"middle": [],
"last": "Hoffart",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Fabian",
"suffix": ""
},
{
"first": "Klaus",
"middle": [],
"last": "Suchanek",
"suffix": ""
},
{
"first": "Gerhard",
"middle": [],
"last": "Berberich",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Weikum",
"suffix": ""
}
],
"year": 2013,
"venue": "Artificial Intelligence",
"volume": "194",
"issue": "",
"pages": "28--61",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Johannes Hoffart, Fabian M Suchanek, Klaus Berberich, and Gerhard Weikum. 2013. Yago2: A spatially and temporally enhanced knowledge base from wikipedia. Artificial Intelligence 194:28-61.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Automatic summarising: The state of the art",
"authors": [
{
"first": "Karen Sp\u00e4rck",
"middle": [],
"last": "Jones",
"suffix": ""
}
],
"year": 2007,
"venue": "Information Processing & Management",
"volume": "43",
"issue": "6",
"pages": "1449--1481",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Karen Sp\u00e4rck Jones. 2007. Automatic summarising: The state of the art. Information Processing & Man- agement 43(6):1449-1481.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Graph summarization with bounded error",
"authors": [
{
"first": "Saket",
"middle": [],
"last": "Navlakha",
"suffix": ""
},
{
"first": "Rajeev",
"middle": [],
"last": "Rastogi",
"suffix": ""
},
{
"first": "Nisheeth",
"middle": [],
"last": "Shrivastava",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of the 2008 ACM SIGMOD international conference on Management of data",
"volume": "",
"issue": "",
"pages": "419--432",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Saket Navlakha, Rajeev Rastogi, and Nisheeth Shri- vastava. 2008. Graph summarization with bounded error. In Proceedings of the 2008 ACM SIGMOD international conference on Management of data. ACM, pages 419-432.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "A survey of text summarization techniques",
"authors": [
{
"first": "Ani",
"middle": [],
"last": "Nenkova",
"suffix": ""
},
{
"first": "Kathleen",
"middle": [],
"last": "Mckeown",
"suffix": ""
}
],
"year": 2012,
"venue": "Mining text data",
"volume": "",
"issue": "",
"pages": "43--76",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ani Nenkova and Kathleen McKeown. 2012. A sur- vey of text summarization techniques. In Mining text data, Springer, pages 43-76.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Monte Carlo statistical methods",
"authors": [
{
"first": "P",
"middle": [],
"last": "Christian",
"suffix": ""
},
{
"first": "George",
"middle": [],
"last": "Robert",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Casella",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "319",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Christian P Robert and George Casella. 2004. Monte Carlo statistical methods, volume 319. Citeseer.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Topic modeling for rdf graphs",
"authors": [
{
"first": "Jennifer",
"middle": [],
"last": "Sleeman",
"suffix": ""
},
{
"first": "Tim",
"middle": [],
"last": "Finin",
"suffix": ""
},
{
"first": "Anupam",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2015,
"venue": "LD4IE@ ISWC",
"volume": "",
"issue": "",
"pages": "48--62",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jennifer Sleeman, Tim Finin, and Anupam Joshi. 2015. Topic modeling for rdf graphs. In LD4IE@ ISWC. pages 48-62.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "A context-aware topic model for statistical machine translation",
"authors": [
{
"first": "Jinsong",
"middle": [],
"last": "Su",
"suffix": ""
},
{
"first": "Deyi",
"middle": [],
"last": "Xiong",
"suffix": ""
},
{
"first": "Yang",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Xianpei",
"middle": [],
"last": "Han",
"suffix": ""
},
{
"first": "Hongyu",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "Junfeng",
"middle": [],
"last": "Yao",
"suffix": ""
},
{
"first": "Min",
"middle": [],
"last": "Zhang",
"suffix": ""
}
],
"year": 2015,
"venue": "ACL (1)",
"volume": "",
"issue": "",
"pages": "229--238",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jinsong Su, Deyi Xiong, Yang Liu, Xianpei Han, Hongyu Lin, Junfeng Yao, and Min Zhang. 2015. A context-aware topic model for statistical machine translation. In ACL (1). pages 229-238.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Linksum: using link analysis to summarize entity data",
"authors": [
{
"first": "Andreas",
"middle": [],
"last": "Thalhammer",
"suffix": ""
},
{
"first": "Nelia",
"middle": [],
"last": "Lasierra",
"suffix": ""
},
{
"first": "Achim",
"middle": [],
"last": "Rettinger",
"suffix": ""
}
],
"year": 2016,
"venue": "International Conference on Web Engineering",
"volume": "",
"issue": "",
"pages": "244--261",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andreas Thalhammer, Nelia Lasierra, and Achim Ret- tinger. 2016. Linksum: using link analysis to sum- marize entity data. In International Conference on Web Engineering. Springer, pages 244-261.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Browsing dbpedia entities with summaries",
"authors": [
{
"first": "Andreas",
"middle": [],
"last": "Thalhammer",
"suffix": ""
},
{
"first": "Achim",
"middle": [],
"last": "Rettinger",
"suffix": ""
}
],
"year": 2014,
"venue": "European Semantic Web Conference",
"volume": "",
"issue": "",
"pages": "511--515",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andreas Thalhammer and Achim Rettinger. 2014. Browsing dbpedia entities with summaries. In Eu- ropean Semantic Web Conference. Springer, pages 511-515.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Trank: Ranking entity types using the web of data",
"authors": [
{
"first": "Alberto",
"middle": [],
"last": "Tonon",
"suffix": ""
},
{
"first": "Michele",
"middle": [],
"last": "Catasta",
"suffix": ""
},
{
"first": "Gianluca",
"middle": [],
"last": "Demartini",
"suffix": ""
},
{
"first": "Philippe",
"middle": [],
"last": "Cudr\u00e9-Mauroux",
"suffix": ""
},
{
"first": "Karl",
"middle": [],
"last": "Aberer",
"suffix": ""
}
],
"year": 2013,
"venue": "International Semantic Web Conference",
"volume": "",
"issue": "",
"pages": "640--656",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alberto Tonon, Michele Catasta, Gianluca Demartini, Philippe Cudr\u00e9-Mauroux, and Karl Aberer. 2013. Trank: Ranking entity types using the web of data. In International Semantic Web Conference. Springer, pages 640-656.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Automatic labeling of topic models using text summaries",
"authors": [
{
"first": "Xiaojun",
"middle": [],
"last": "Wan",
"suffix": ""
},
{
"first": "Tianming",
"middle": [],
"last": "Wang",
"suffix": ""
}
],
"year": 2016,
"venue": "ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xiaojun Wan and Tianming Wang. 2016. Automatic labeling of topic models using text summaries. In ACL (1).",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "A biterm topic model for short texts",
"authors": [
{
"first": "Xiaohui",
"middle": [],
"last": "Yan",
"suffix": ""
},
{
"first": "Jiafeng",
"middle": [],
"last": "Guo",
"suffix": ""
},
{
"first": "Yanyan",
"middle": [],
"last": "Lan",
"suffix": ""
},
{
"first": "Xueqi",
"middle": [],
"last": "Cheng",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the 22nd international conference on World Wide Web",
"volume": "",
"issue": "",
"pages": "1445--1456",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xiaohui Yan, Jiafeng Guo, Yanyan Lan, and Xueqi Cheng. 2013. A biterm topic model for short texts. In Proceedings of the 22nd international conference on World Wide Web. ACM, pages 1445-1456.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Ontology summarization based on rdf sentence graph",
"authors": [
{
"first": "Xiang",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Gong",
"middle": [],
"last": "Cheng",
"suffix": ""
},
{
"first": "Yuzhong",
"middle": [],
"last": "Qu",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 16th international conference on World Wide Web",
"volume": "",
"issue": "",
"pages": "707--716",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xiang Zhang, Gong Cheng, and Yuzhong Qu. 2007. Ontology summarization based on rdf sentence graph. In Proceedings of the 16th international con- ference on World Wide Web. ACM, pages 707-716.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Unsupervised segmentation helps supervised learning of character tagging for word segmentation and named entity recognition",
"authors": [
{
"first": "Hai",
"middle": [],
"last": "Zhao",
"suffix": ""
},
{
"first": "Chunyu",
"middle": [],
"last": "Kit",
"suffix": ""
}
],
"year": 2008,
"venue": "IJCNLP",
"volume": "",
"issue": "",
"pages": "106--111",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hai Zhao and Chunyu Kit. 2008. Unsupervised seg- mentation helps supervised learning of character tagging for word segmentation and named entity recognition. In IJCNLP. pages 106-111.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"text": "Entity Summarization Model mon for triples to contain unexpected characters.",
"num": null,
"type_str": "figure",
"uris": null
},
"TABREF0": {
"content": "<table><tr><td>5</td><td>Draw a predicate distribution \u03b8 d \u223c</td></tr><tr><td/><td>Dir(\u03b1 d )</td></tr><tr><td>6</td><td>foreach subject s and object o of</td></tr><tr><td/><td>document d do</td></tr><tr><td>7</td><td>Draw a predicate r \u223c Mult(\u03b8 d )</td></tr><tr><td>8</td><td>Draw a subject s from predicate</td></tr><tr><td/><td>r, s \u223c Mult(\u03c6 r )</td></tr><tr><td>9</td><td>Draw an object o from predicate</td></tr><tr><td/><td>r, o \u223c Mult(\u03c6 r )</td></tr><tr><td>10</td><td>end</td></tr><tr><td colspan=\"2\">11 end</td></tr></table>",
"text": "Algorithm 1: ES-LDA Model1 foreach predicate r \u2208 {1, 2, . . . , R} do 2 Draw an object distribution \u03c6 r \u223c Dir(\u03b2 r \u00d7 \u039b r ) 3 end 4 foreach document d \u2208 {1, 2, . . . , D} do",
"num": null,
"type_str": "table",
"html": null
},
"TABREF1": {
"content": "<table/>",
"text": "Overall quality results of different models. Best result are bold.",
"num": null,
"type_str": "table",
"html": null
},
"TABREF2": {
"content": "<table><tr><td/><td>MARIE CURIE</td><td/><td/><td colspan=\"2\">REIGN OF FIRE</td><td/><td>SEYCHELLES</td><td/></tr><tr><td>ES-LDA</td><td>LinkSUM</td><td>FACES</td><td>ES-LDA</td><td colspan=\"2\">LinkSUM FACES</td><td>ES-LDA</td><td>LinkSUM</td><td>FACES</td></tr><tr><td colspan=\"2\">doctoralStudents birthPlace</td><td>spouse</td><td>starring</td><td>country</td><td>starring</td><td>leaderName</td><td>largestCity</td><td>leaderName</td></tr><tr><td colspan=\"2\">doctoralAdvisor birthPlace</td><td>field</td><td>producer</td><td>starring</td><td>country</td><td colspan=\"3\">governmentType governmentType governmentType</td></tr><tr><td>deathPlace</td><td>field</td><td colspan=\"2\">workInstitutions music</td><td>starring</td><td>distributor</td><td>leaderTitle</td><td colspan=\"2\">governmentType largestCity</td></tr><tr><td>children</td><td>field</td><td>birthPlace</td><td>director</td><td>starring</td><td colspan=\"4\">musicComposer officialLanguage governmentType sovereigntyType</td></tr><tr><td>knownFor</td><td>knownFor</td><td>deathPlace</td><td colspan=\"2\">cinematography studio</td><td>director</td><td>capital</td><td colspan=\"2\">governmentType source</td></tr><tr><td>spouse</td><td>almaMater</td><td colspan=\"2\">doctoralAdvisor country</td><td>producer</td><td>editing</td><td>currency</td><td colspan=\"2\">sovereigntyType capital</td></tr><tr><td>almaMater</td><td>birthPlace</td><td>knownFor</td><td>distributor</td><td>producer</td><td>studio</td><td>timeZone</td><td>source</td><td>leaderTitle</td></tr><tr><td>birthPlace</td><td>knownFor</td><td>almaMater</td><td>studio</td><td>director</td><td>music</td><td>legislature</td><td>capital</td><td>language</td></tr><tr><td>field</td><td colspan=\"3\">doctoralAdvisor doctoralStudents editing</td><td>artist</td><td>producer</td><td>anthem</td><td>language</td><td>languages</td></tr><tr><td colspan=\"2\">establishedEvent knownFor</td><td>thumbnail</td><td>screenplay</td><td>producer</td><td>thumbnail</td><td>callingCode</td><td>timeZone</td><td>legislature</td></tr></table>",
"text": "Top-10 predicates for three randomly selected entities after applying three different models.",
"num": null,
"type_str": "table",
"html": null
},
"TABREF3": {
"content": "<table><tr><td colspan=\"2\">LEXUS</td><td colspan=\"2\">MORTAL KOMBAT TRILOGY</td></tr><tr><td>Predicate</td><td colspan=\"2\">Probability Predicate</td><td>Probability</td></tr><tr><td colspan=\"2\">foundedBy 0.21</td><td>platforms</td><td>0.30</td></tr><tr><td>owner</td><td>0.17</td><td>publisher</td><td>0.18</td></tr><tr><td>location</td><td>0.15</td><td>developer</td><td>0.17</td></tr><tr><td colspan=\"2\">keyPerson 0.06</td><td colspan=\"2\">computingMedia 0.07</td></tr><tr><td>service</td><td>0.04</td><td>designer</td><td>0.05</td></tr></table>",
"text": "Probabilities of top-5 predicates for two randomly selected entities.",
"num": null,
"type_str": "table",
"html": null
},
"TABREF4": {
"content": "<table><tr><td>PARTY</td><td/><td>STARRING</td><td/></tr><tr><td>Object</td><td colspan=\"2\">Probability Object</td><td>Probability</td></tr><tr><td>Democratic Party (United States)</td><td>0.36</td><td colspan=\"2\">Arnold Schwarzenegger 0.05</td></tr><tr><td>Republican Party (United States)</td><td>0.17</td><td>Angelina Jolie</td><td>0.04</td></tr><tr><td>Democratic-Republican Party</td><td>0.12</td><td>Raven Symone</td><td>0.03</td></tr><tr><td colspan=\"2\">Communist Party of the Soviet Union 0.08</td><td colspan=\"2\">Matthew McConaughey 0.02</td></tr><tr><td>Independent(politician)</td><td>0.08</td><td>Alan Arkin</td><td>0.02</td></tr></table>",
"text": "Distributions of two randomly selected predicates over top-5 objects.",
"num": null,
"type_str": "table",
"html": null
}
}
}
} |