File size: 99,308 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 |
{
"paper_id": "I11-1024",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T07:31:26.856940Z"
},
"title": "An Empirical Study on Compositionality in Compound Nouns",
"authors": [
{
"first": "Siva",
"middle": [],
"last": "Reddy",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of York",
"location": {
"country": "UK"
}
},
"email": ""
},
{
"first": "Diana",
"middle": [],
"last": "Mccarthy",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Lexical Computing Ltd",
"location": {
"country": "UK"
}
},
"email": ""
},
{
"first": "Suresh",
"middle": [],
"last": "Manandhar",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of York",
"location": {
"country": "UK"
}
},
"email": "suresh@cs.york.ac.uk"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "A multiword is compositional if its meaning can be expressed in terms of the meaning of its constituents. In this paper, we collect and analyse the compositionality judgments for a range of compound nouns using Mechanical Turk. Unlike existing compositionality datasets, our dataset has judgments on the contribution of constituent words as well as judgments for the phrase as a whole. We use this dataset to study the relation between the judgments at constituent level to that for the whole phrase. We then evaluate two different types of distributional models for compositionality detection-constituent based models and composition function based models. Both the models show competitive performance though the composition function based models perform slightly better. In both types, additive models perform better than their multiplicative counterparts.",
"pdf_parse": {
"paper_id": "I11-1024",
"_pdf_hash": "",
"abstract": [
{
"text": "A multiword is compositional if its meaning can be expressed in terms of the meaning of its constituents. In this paper, we collect and analyse the compositionality judgments for a range of compound nouns using Mechanical Turk. Unlike existing compositionality datasets, our dataset has judgments on the contribution of constituent words as well as judgments for the phrase as a whole. We use this dataset to study the relation between the judgments at constituent level to that for the whole phrase. We then evaluate two different types of distributional models for compositionality detection-constituent based models and composition function based models. Both the models show competitive performance though the composition function based models perform slightly better. In both types, additive models perform better than their multiplicative counterparts.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Compositionality is a language phenomenon where the meaning of an expression can be expressed in terms of the meaning of its constituents. Multiword expressions (Sag et al., 2002, MWEs) are known to display a continuum of compositionality (McCarthy et al., 2003) where some of them are compositional e.g. \"swimming pool\", some are non-compositional e.g. \"cloud nine\", and some in between e.g. \"zebra crossing\".",
"cite_spans": [
{
"start": 161,
"end": 185,
"text": "(Sag et al., 2002, MWEs)",
"ref_id": null
},
{
"start": 239,
"end": 262,
"text": "(McCarthy et al., 2003)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The past decade has seen interest in developing computational methods for compositionality in MWEs (Lin, 1999; Schone and Jurafsky, 2001; Baldwin et al., 2003; Bannard et al., 2003; Mc-Carthy et al., 2003; Venkatapathy and Joshi, 2005; Katz and Giesbrecht, 2006; Sporleder and Li, 2009) . Recent developments in vector-based semantic composition functions (Mitchell and Lapata, 2008; Widdows, 2008) have also been applied to compositionality detection (Giesbrecht, 2009) .",
"cite_spans": [
{
"start": 99,
"end": 110,
"text": "(Lin, 1999;",
"ref_id": "BIBREF12"
},
{
"start": 111,
"end": 137,
"text": "Schone and Jurafsky, 2001;",
"ref_id": "BIBREF22"
},
{
"start": 138,
"end": 159,
"text": "Baldwin et al., 2003;",
"ref_id": "BIBREF0"
},
{
"start": 160,
"end": 181,
"text": "Bannard et al., 2003;",
"ref_id": "BIBREF0"
},
{
"start": 182,
"end": 205,
"text": "Mc-Carthy et al., 2003;",
"ref_id": null
},
{
"start": 206,
"end": 235,
"text": "Venkatapathy and Joshi, 2005;",
"ref_id": "BIBREF26"
},
{
"start": 236,
"end": 262,
"text": "Katz and Giesbrecht, 2006;",
"ref_id": "BIBREF9"
},
{
"start": 263,
"end": 286,
"text": "Sporleder and Li, 2009)",
"ref_id": "BIBREF24"
},
{
"start": 356,
"end": 383,
"text": "(Mitchell and Lapata, 2008;",
"ref_id": "BIBREF16"
},
{
"start": 384,
"end": 398,
"text": "Widdows, 2008)",
"ref_id": "BIBREF27"
},
{
"start": 452,
"end": 470,
"text": "(Giesbrecht, 2009)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "While the existing methods of compositionality detection use constituent word level semantics to compose the semantics of the phrase, the evaluation datasets are not particularly suitable to study the contribution of each constituent word to the semantics of the phrase. Existing datasets (McCarthy et al., 2003; Venkatapathy and Joshi, 2005; Katz and Giesbrecht, 2006; Biemann and Giesbrecht, 2011) only have the compositionality judgment of the whole expression without constituent word level judgment, or they have judgments on the constituents without judgments on the whole (Bannard et al., 2003) . Our dataset allows us to examine the relationship between the two rather than assume the nature of it.",
"cite_spans": [
{
"start": 289,
"end": 312,
"text": "(McCarthy et al., 2003;",
"ref_id": "BIBREF13"
},
{
"start": 313,
"end": 342,
"text": "Venkatapathy and Joshi, 2005;",
"ref_id": "BIBREF26"
},
{
"start": 343,
"end": 369,
"text": "Katz and Giesbrecht, 2006;",
"ref_id": "BIBREF9"
},
{
"start": 370,
"end": 399,
"text": "Biemann and Giesbrecht, 2011)",
"ref_id": "BIBREF1"
},
{
"start": 579,
"end": 601,
"text": "(Bannard et al., 2003)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this paper we collect judgments of the contribution of constituent nouns within noun-noun compounds (section 2) alongside judgments of compositionality of the compound. We study the relation between the contribution of the parts with the compositionality of the whole (section 3). We propose various constituent based models (section 4.3) which are intuitive and related to existing models of compositionality detection (section 4.1) and we evaluate these models in comparison to composition function based models. All the models discussed in this paper are built using a distributional word-space model approach (Sahlgren, 2006) .",
"cite_spans": [
{
"start": 616,
"end": 632,
"text": "(Sahlgren, 2006)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this section, we describe the experimental setup for the collecting compositionality judgments of English compound nouns. All the existing datasets focused either on verb-particle, verbnoun or adjective-noun phrases. Instead, we focus on compound nouns for which resources are rel-atively scarce. In this paper, we only deal with compound nouns made up of two words separated by space.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compositionality in Compound Nouns",
"sec_num": "2"
},
{
"text": "In the literature (Nunberg et al., 1994; Baldwin et al., 2003; Fazly et al., 2009) , compositionality is discussed in many terms including simple decomposable, semantically analyzable, idiosyncratically decomposable and non-decomposable. For practical NLP purposes, Bannard et al. (2003) adopt a straightforward definition of a compound being compositional if \"the overall semantics of the multi-word expression (here compound) can be composed from the simplex semantics of its parts, as described (explicitly or implicitly) in a finite lexicon\". We adopt this definition and pose compositionality as a literality issue. A compound is compositional if its meaning can be understood from the literal (simplex) meaning of its parts. Similar views of compositionality as literality are found in (Lin, 1999; Katz and Giesbrecht, 2006) . In the past there have been arguments in favor/disfavor of compositionality as literality approach (e.g. see (Gibbs, 1989; Titone and Connine, 1999) ). The idea of viewing compositionality as literality is also motivated from the shared task organized by Biemann and Giesbrecht (2011) . From here on, we use the terms compositionality and literality interchangeably.",
"cite_spans": [
{
"start": 18,
"end": 40,
"text": "(Nunberg et al., 1994;",
"ref_id": null
},
{
"start": 41,
"end": 62,
"text": "Baldwin et al., 2003;",
"ref_id": "BIBREF0"
},
{
"start": 63,
"end": 82,
"text": "Fazly et al., 2009)",
"ref_id": "BIBREF3"
},
{
"start": 266,
"end": 287,
"text": "Bannard et al. (2003)",
"ref_id": "BIBREF0"
},
{
"start": 792,
"end": 803,
"text": "(Lin, 1999;",
"ref_id": "BIBREF12"
},
{
"start": 804,
"end": 830,
"text": "Katz and Giesbrecht, 2006)",
"ref_id": "BIBREF9"
},
{
"start": 942,
"end": 955,
"text": "(Gibbs, 1989;",
"ref_id": "BIBREF6"
},
{
"start": 956,
"end": 981,
"text": "Titone and Connine, 1999)",
"ref_id": "BIBREF25"
},
{
"start": 1088,
"end": 1117,
"text": "Biemann and Giesbrecht (2011)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation setup",
"sec_num": "2.1"
},
{
"text": "We ask humans to score the compositionality of a phrase by asking them how literal the phrase is. Since we wish to see in our data the extent that the phrase is compositional, and to what extent that depends on the contribution in meaning of its parts, we also ask them how literal the use of a component word is within the given phrase.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation setup",
"sec_num": "2.1"
},
{
"text": "For each compound noun, we create three separate tasks -one for each constituent's literality and one for the phrase compositionality. The motivation behind using three separate tasks is to make the scoring mechanism for each task independent of the other tasks. This enables us to study the actual relation between the constituents and the compound scores without any bias to any particular annotator's way of arriving at the scores of the compound w.r.t. the constituents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation setup",
"sec_num": "2.1"
},
{
"text": "There are many factors to consider in eliciting compositionality judgments, such as ambiguity of the expression and individual variation of annotator in background knowledge. To control for this, we ask subjects if they can interpret the meaning of a compound noun from only the meaning of the component nouns where we also provide contextual information. All the possible definitions of a compound noun are chosen from WordNet (Fellbaum, 1998) , Wiktionary or defined by ourselves if some of the definitions are absent. Five examples of each compound noun are randomly chosen from the ukWaC (Ferraresi et al., 2008) corpus and the same set of examples are displayed to all the annotators. The annotators select the definition of the compound noun which occurs most frequently in the examples and then score the compound for literality based on the most frequent definition.",
"cite_spans": [
{
"start": 428,
"end": 444,
"text": "(Fellbaum, 1998)",
"ref_id": null
},
{
"start": 592,
"end": 616,
"text": "(Ferraresi et al., 2008)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation setup",
"sec_num": "2.1"
},
{
"text": "We have two reasons for making the annotators read the examples, choose the most frequent definition and base literality judgments on the most frequent definition. The first reason is to provide a context to the decisions and reduce the impact of ambiguity. The second is that distributional models are greatly influenced by frequency and since we aim to work with distributional models for compositionality detection we base our findings on the most frequent sense of the compound noun. In this work we consider the compositionality of the noun-noun compound type without token based disambiguation which we leave for future work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Annotation setup",
"sec_num": "2.1"
},
{
"text": "We could not find any compound noun datasets publicly available which are marked for compositionality judgments. Korkontzelos and Manandhar (2009) prepared a related dataset for compound nouns but compositionality scores were absent and their set contains only 38 compounds. There are datasets for verb-particle (McCarthy et al., 2003) , verb-noun judgments (Biemann and Giesbrecht, 2011; Venkatapathy and Joshi, 2005) and adjective-noun (Biemann and Giesbrecht, 2011) . Not only are these not the focus of our work, but also we wanted datasets with each constituent word's literality score. Bannard et al. (2003) obtained judgments on whether a verb-particle construction implies the verb or the particle or both. The judgments were binary and not on a scale and there was no judgment of compositionality of the whole construction. Ours is the first attempt to provide a dataset which have both scalar compositionality judgments of the phrase as well as the literality score for each component word.",
"cite_spans": [
{
"start": 113,
"end": 146,
"text": "Korkontzelos and Manandhar (2009)",
"ref_id": "BIBREF10"
},
{
"start": 312,
"end": 335,
"text": "(McCarthy et al., 2003)",
"ref_id": "BIBREF13"
},
{
"start": 358,
"end": 388,
"text": "(Biemann and Giesbrecht, 2011;",
"ref_id": "BIBREF1"
},
{
"start": 389,
"end": 418,
"text": "Venkatapathy and Joshi, 2005)",
"ref_id": "BIBREF26"
},
{
"start": 438,
"end": 468,
"text": "(Biemann and Giesbrecht, 2011)",
"ref_id": "BIBREF1"
},
{
"start": 592,
"end": 613,
"text": "Bannard et al. (2003)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "We aimed for a dataset which would include compound nouns where: 1) both the component words are used literally, 2) the first word is used literally but not the second, 3) the second word is used literally but not the first and 4) both the words are used non-literally. Such a dataset would provide stronger evidence to study the relation between the constituents of the compound noun and its compositionality behaviour.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "We used the following heuristics based on WordNet to classify compound nouns into 4 above classes.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "1. Each of the component word exists either in the hypernymy hierarchy of the compound noun or in the definition(s) of the compound noun. e.g. swimming pool because swimming exists in the WordNet definition of swimming pool and pool exists in the hypernymy hierarchy of swimming pool 2. Only the first word exists either in the hypernymy hierarchy or in the definition(s) of the compound and not the second word. e.g. night owl 3. Only the second word exists either in the hypernymy hierarchy or in the definition(s) of the compound and not the first word. e.g. zebra crossing 4. Neither of the words exist either in hypernymy hierarchy or in the definition(s) of the compound noun. e.g. smoking gun",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "The intuition behind the heuristics is that if a component word is used literally in a compound, it would probably be used in the definition of the compound or may appear in the synset hierarchy of the compound. We changed the constraints, for example decreasing/increasing the depth of the hypernymy hierarchy, and for each class we randomly picked 30 potential candidates by rough manual verification. There were fewer instances in the classes 2 and 4. In order to populate these classes, we selected additional compound nouns from Wiktionary by manually inspecting if they can fall in either class.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "These heuristics were only used for obtaining our sample, they were not used for categorizing the compound nouns in our study. The compound nouns in all these temporary classes are merged and 90 compound words are selected which have at least 50 instances in the ukWaC corpus. These 90 compound words are chosen for the dataset. Snow et al. (2008) used Amazon mechanical turk (AMT) for annotating language processing tasks. They found that although an individual turker (annotator) performance was lower compared to an expert, as the number of turkers increases, the quality of the annotated data surpassed expert level quality. We used 30 turkers for annotating each single task and then retained the judgments with sufficient consensus as described in section 2.4.",
"cite_spans": [
{
"start": 329,
"end": 347,
"text": "Snow et al. (2008)",
"ref_id": "BIBREF23"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Compound noun dataset",
"sec_num": "2.2"
},
{
"text": "For each compound noun, 3 types of tasks are created as described above: a judgment on how literal the phrase is and a judgment on how literal each noun is within the compound. For 90 compound nouns, 270 independent tasks are therefore created. Each of these tasks is assigned to 30 annotators. A task is assigned randomly to an annotator by AMT so each annotator may work on only some of the tasks for a given compound.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Annotators",
"sec_num": "2.3"
},
{
"text": "Recent studies 1 shows that AMT data is prone to spammers and outliers. We dealt with them in three ways. a). We designed a qualification test 2 which provides an annotator with basic training about literality, and they can participate in the annotation task only if they pass the test. b). Once all the annotations (90 phrases * 3 tasks/phrase * 30 annotations/task = 8100 annotations) are completed, we calculated the average Spearman correlation score (\u03c1) of every annotator by correlating their annotation values with every other annotator and taking the average. We discarded the work of annotators whose \u03c1 is negative and accepted all the work of annotators whose \u03c1 is greater than 0.6. c). For the other annotators, we accepted their annotation for a task only if their annotation judgment is within the range of \u00b11.5 from the task's mean. Table 1 displays AMT statistics. Overall, each annotator on average worked on 53 tasks randomly selected from the set of 270 tasks. This lowers the chance of bias in the data because of any particular annotator.",
"cite_spans": [],
"ref_spans": [
{
"start": 847,
"end": 854,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Quality of the annotations",
"sec_num": "2.4"
},
{
"text": "Spearman correlation scores \u03c1 provide an estimate of annotator agreement. To know the difficulty level of the three types of tasks described in section 2, \u03c1 for each task type is also displayed in For each compound, we also studied the distribution of scores around the mean by observing the standard deviation \u03c3. All the compound nouns along with their mean and standard deviations are shown in table 2.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Quality of the annotations",
"sec_num": "2.4"
},
{
"text": "Ideally, if all the annotators agree on a judgment for a given compound or a component word, the deviation should be low. Among the 90 compounds, 15 of them are found to have a deviation > \u00b11.5. We used this threshold to signify annotator disagreement. The reason for disagreement could be due to the ambiguity of the compound e.g. silver screen, brass ring or due to the subjective differences of opinion between the annotators.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Quality of the annotations",
"sec_num": "2.4"
},
{
"text": "Overall, the inter annotator agreement (\u03c1) is high and the standard deviation of most tasks is low (except for a few exceptions). So we are confident that the dataset can be used as a reliable goldstandard with which we conduct experiments. The dataset is publicly available for download 3 . 3 Analyzing the Human Judgments",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Quality of the annotations",
"sec_num": "2.4"
},
{
"text": "By analyzing the mean values of the phrase level annotations, we found that compounds displayed a varied level of compositionality. For some compounds annotators confirm that they can interpret the meaning of a compound from its component words and for some they do not. For others they grade in-between. Figure 1 displays the mean values of compositionality scores of all compounds. Compounds are arranged along the X-axis in increasing order of their score. The graph displays a continuum of compositionality (McCarthy et al., 2003) . We note that our sample of compounds was selected to exhibit a range of compositionality.",
"cite_spans": [
{
"start": 511,
"end": 534,
"text": "(McCarthy et al., 2003)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [
{
"start": 305,
"end": 313,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Quality of the annotations",
"sec_num": "2.4"
},
{
"text": "The dataset allows us to study the relation between constituent word level contributions to the phrase level compositionality scores. Let w1 and w2 be the constituent words of the compound w3. Let s1, s2 and s3 be the mean literality scores of w1, w2 and w3 respectively. Using a 3-fold cross validation on the annotated data, we tried various function fittings f over the judgments s1, s2 and s3.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "\u2022 ADD: a.s1 + b.s2 = s3",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "\u2022 MULT: a.s1.s2 = s3 \u2022 COMB: a.s1 + b.s2 + c.s1.s2 = s3 \u2022 WORD1: a.s1 = s3 \u2022 WORD2: a.s2 = s3",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "where a, b and c are coefficients.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "We performed 3-fold cross validation to evaluate the above functions (two training samples and Results (both \u03c1 and R 2 ) clearly show that a relation exists between the constituent literality scores and the phrase compositionality. Existing compositionality approaches on noun-noun compounds such as (Baldwin et al., 2003; Korkontzelos and Manandhar, 2009) use the semantics of only one of the constituent words (generally the head word)",
"cite_spans": [
{
"start": 300,
"end": 322,
"text": "(Baldwin et al., 2003;",
"ref_id": "BIBREF0"
},
{
"start": 323,
"end": 356,
"text": "Korkontzelos and Manandhar, 2009)",
"ref_id": "BIBREF10"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "to determine the compositionality of the phrase. But the goodness of fit R 2 values show that the functions ADD, COMB and MULT which intuitively make use of both the constituent scores fit the data better than functions using only one of the constituents. Furthermore, COMB and ADD suggest that additive models are preferable to multiplicative. In this data, the first constituent word plays a slightly more important role than the second in determining compositionality.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "Overall, this study suggests that it is possible to estimate the phrase level compositionality scores given the constituent word level literality scores. This motivates us to present constituent based models (section 4.3) for compositionality score estimation of a compound. We begin the next section on computational models with a discussion of related work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Relation between the constituents and the phrase compositionality judgments",
"sec_num": "3.1"
},
{
"text": "Most methods in compositionality detection can be classified into two types -those which make use of lexical fixedness and syntactic properties of the MWEs, and those which make use of the semantic similarities between the constituents and the MWE.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "4.1"
},
{
"text": "Non compositional MWEs are known to have lexical fixedness in which the component words have high statistical association. Some of the methods which exploit this feature are (Lin, 1999; Pedersen, 2011) . This property does not hold always because institutionalized MWEs (Sag et al., 2002) are known to have high association even though they are compositional, especially in the case of compound nouns. Another property of non-compositional MWEs is that they show syntactic rigidness which do not allow internal modifiers or morphological variations of the components, or variations that break typical selectional preferences. Methods like (Cook et al., 2007; Mc-Carthy et al., 2007; Fazly et al., 2009 ) exploit this property. This holds mostly for verbal idioms but not for compound nouns since the variations of any compound noun are highly limited.",
"cite_spans": [
{
"start": 174,
"end": 185,
"text": "(Lin, 1999;",
"ref_id": "BIBREF12"
},
{
"start": 186,
"end": 201,
"text": "Pedersen, 2011)",
"ref_id": "BIBREF18"
},
{
"start": 265,
"end": 288,
"text": "MWEs (Sag et al., 2002)",
"ref_id": null
},
{
"start": 639,
"end": 658,
"text": "(Cook et al., 2007;",
"ref_id": "BIBREF2"
},
{
"start": 659,
"end": 682,
"text": "Mc-Carthy et al., 2007;",
"ref_id": null
},
{
"start": 683,
"end": 701,
"text": "Fazly et al., 2009",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "4.1"
},
{
"text": "Other methods like (Baldwin et al., 2003; Sporleder and Li, 2009) are based on semantic similarities between the constituents and the MWE. Baldwin et al. (2003) use only the information of the semantic similarity between one of the constituents and the compound to determine the compositionality. Sporleder and Li (2009) determine the compositionality of verbal phrases in a given context (token-based disambiguation) based on the lexical chain similarities of the constituents and the context of the MWE. Bannard et al. (2003) and McCarthy et al. (2003) study the compositionality in verb particles and they found that methods based on the similarity between simplex parts (constituents) and the phrases are useful to study semantics of the phrases. These findings motivated our constituent based models along with the findings in section 3.1.",
"cite_spans": [
{
"start": 19,
"end": 41,
"text": "(Baldwin et al., 2003;",
"ref_id": "BIBREF0"
},
{
"start": 42,
"end": 65,
"text": "Sporleder and Li, 2009)",
"ref_id": "BIBREF24"
},
{
"start": 139,
"end": 160,
"text": "Baldwin et al. (2003)",
"ref_id": "BIBREF0"
},
{
"start": 297,
"end": 320,
"text": "Sporleder and Li (2009)",
"ref_id": "BIBREF24"
},
{
"start": 506,
"end": 527,
"text": "Bannard et al. (2003)",
"ref_id": "BIBREF0"
},
{
"start": 532,
"end": 554,
"text": "McCarthy et al. (2003)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "4.1"
},
{
"text": "In addition to the constituent based models (section 4.3), there are composition function based vector models (Mitchell and Lapata, 2008; Widdows, 2008) which make use of the semantics of the constituents in a different manner. These models are described in section 4.4 and are evaluated in comparison with the constituent-based models.",
"cite_spans": [
{
"start": 110,
"end": 137,
"text": "(Mitchell and Lapata, 2008;",
"ref_id": "BIBREF16"
},
{
"start": 138,
"end": 152,
"text": "Widdows, 2008)",
"ref_id": "BIBREF27"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "4.1"
},
{
"text": "The vector space model used in all our experiments is described as follows.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Related work",
"sec_num": "4.1"
},
{
"text": "Our vector space model is also called a word space model (Sahlgren, 2006 , WSM) since we represent a word's meaning in a dimensional space. In the WSM, a word meaning is represented in terms of its Co-occurrences observed in a large corpora where the co-occurrences are stored in a vector format.",
"cite_spans": [
{
"start": 57,
"end": 72,
"text": "(Sahlgren, 2006",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Vector space model of meaning",
"sec_num": "4.2"
},
{
"text": "The lemmatised context words around the target word in a window of size 100 are treated as the co-occurrences. The top 10000 frequent content words in the ukWaC (along with their part-of-speech category) are used for the feature co-occurrences i.e. the dimensionality of the WSM. To measure similarity between two vectors, cosine similarity (sim) is used. Following Mitchell and Lapata (2008) , the context words in the vector are set to the ratio of probability of the context word given the target word to the overall probability of the context word 4 .",
"cite_spans": [
{
"start": 366,
"end": 392,
"text": "Mitchell and Lapata (2008)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Vector space model of meaning",
"sec_num": "4.2"
},
{
"text": "Given a compound word w3 with the constituents w1 and w2, constituent based models determine the compositionality score s3 of the compound by first determining the literality scores s1 and s2 of w1 and w2 respectively (section 4.3.1) and then using one of the functions f (described in section 3.1), the compositionality score s3 is estimated using s3 = f (s1, s2) (section 4.3.2).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Constituent based models",
"sec_num": "4.3"
},
{
"text": "If a constituent word is used literally in a given compound it is highly likely that the compound and the constituent share common co-occurrences. For example, the compound swimming pool has the co-occurrences water, fun and indoor which are also commonly found with the constituents swimming and pool.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Literality scores of the constituents",
"sec_num": "4.3.1"
},
{
"text": "We define the literality of a word in a given compound as the similarity between the compound and the constituent co-occurrence vectors i.e. if the number of common co-occurrences are numerous then the constituent is more likely to be meant literally in the compound.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Literality scores of the constituents",
"sec_num": "4.3.1"
},
{
"text": "Let v1, v2 and v3 be the co-occurrence vectors of w1, w2 and w3. The literality scores s1 and s2 of w1 and w2 in the compound w3 are defined as s1 = sim(v1, v3) s2 = sim (v2, v3) where sim is the cosine similarity between the vectors.",
"cite_spans": [
{
"start": 170,
"end": 178,
"text": "(v2, v3)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Literality scores of the constituents",
"sec_num": "4.3.1"
},
{
"text": "Given the literality scores s1 and s2 of the constituents, we can now compute the compositionality score s3 of the compound w3 using any of the functions f defined in section 3.1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compositionality of the compound",
"sec_num": "4.3.2"
},
{
"text": "s3 = f (s1, s2)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Compositionality of the compound",
"sec_num": "4.3.2"
},
{
"text": "In these models (Schone and Jurafsky, 2001; Katz and Giesbrecht, 2006; Giesbrecht, 2009) of compositionality detection, firstly a vector for the compound is composed from its constituents using a compositionality function \u2295. Then the similarity between the composed vector and true cooccurrence vector of the compound is measured to determine the compositionality: the higher the similarity, the higher the compositionality of the compound. Guevara (2011) observed that additive models performed well for building composition vectors of phrases from their parts whereas Mitchell and Lapata (2008) found in favor of multiplicative models. We experiment using both the compositionality functions simple addition 5 and simple multiplication, which are the most widely used composition functions, known for their simplicity and good performance.",
"cite_spans": [
{
"start": 16,
"end": 43,
"text": "(Schone and Jurafsky, 2001;",
"ref_id": "BIBREF22"
},
{
"start": 44,
"end": 70,
"text": "Katz and Giesbrecht, 2006;",
"ref_id": "BIBREF9"
},
{
"start": 71,
"end": 88,
"text": "Giesbrecht, 2009)",
"ref_id": "BIBREF7"
},
{
"start": 570,
"end": 596,
"text": "Mitchell and Lapata (2008)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Composition function based models",
"sec_num": "4.4"
},
{
"text": "Vector v1 \u2295 v2 for a compound w 3 is composed from its constituent word vectors v1 and v2 using the vector addition av1 + bv2 and simple multiplication v1v2 where the i th element of v1 \u2295 v2 is defined as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition function based models",
"sec_num": "4.4"
},
{
"text": "(av1 + bv2) i = a.v1 i + b.v2 i (v1v2) i = v1 i .v2 i",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition function based models",
"sec_num": "4.4"
},
{
"text": "first constituent second constituent s1 0.616 -s2 -0.707 Table 4 : Constituent level correlations",
"cite_spans": [],
"ref_spans": [
{
"start": 57,
"end": 64,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Composition function based models",
"sec_num": "4.4"
},
{
"text": "The compositionality score of the compound is then measured using s3 = sim(v1\u2295v2, v3) where v3 is the co-occurrence vector of the compound built from the corpus. For more details of these models please refer to (Mitchell and Lapata, 2008; Giesbrecht, 2009) .",
"cite_spans": [
{
"start": 211,
"end": 238,
"text": "(Mitchell and Lapata, 2008;",
"ref_id": "BIBREF16"
},
{
"start": 239,
"end": 256,
"text": "Giesbrecht, 2009)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Composition function based models",
"sec_num": "4.4"
},
{
"text": "We evaluated all the models on the dataset developed in section 2. Since our dataset has constituent level contributions along with phrase compositionality judgments, we evaluated the constituent based models against both the literality scores of the constituents (section 4.3.1) and the phrase level judgments (section 4.3.2). The composition function models are evaluated only on phrase level scores following (McCarthy et al., 2003; Venkatapathy and Joshi, 2005; Biemann and Giesbrecht, 2011) : higher correlation scores indicate better compositionality predictions.",
"cite_spans": [
{
"start": 412,
"end": 435,
"text": "(McCarthy et al., 2003;",
"ref_id": "BIBREF13"
},
{
"start": 436,
"end": 465,
"text": "Venkatapathy and Joshi, 2005;",
"ref_id": "BIBREF26"
},
{
"start": 466,
"end": 495,
"text": "Biemann and Giesbrecht, 2011)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation",
"sec_num": "4.5"
},
{
"text": "Spearman's \u03c1 correlations of s1 and s2 with the human constituent level judgments are shown in table 4. We observed that the predictions for the second constituent are more accurate than those for the first constituent. Perhaps these constitute an easier set of nouns for modelling but we need to investigate this further.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Constituent based models evaluation",
"sec_num": null
},
{
"text": "For the phrase compositionality evaluation we did a 3-fold cross validation. The parameters of the functions f (section 4.3.2) are predicted by least square linear regression over the training samples and optimum values are selected. The average Spearman correlation scores of phrase compositionality scores with human judgements on the testing samples are displayed in table 5. The goodness of fit R 2 values when trained over the whole dataset are also displayed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Constituent based models evaluation",
"sec_num": null
},
{
"text": "It is clear that models ADD and COMB which use both the constituents are better predictors of phrase compositionality compared to the single word based predictors WORD1 and WORD2. Both ADD and COMB are competitive in terms of both the correlations (accuracy) and goodness of Model sim(v1, v3) ) was found to be a better predictor of phrase compositionality than the second (WORD2) following the behaviour of the mechanical turkers as in table 3.",
"cite_spans": [
{
"start": 281,
"end": 292,
"text": "sim(v1, v3)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Constituent based models evaluation",
"sec_num": null
},
{
"text": "These models are evaluated for phrase compositionality scores. As with the constituent based models, for estimating the model parameters a and b of the composition function based models, we did a 3-fold cross validation. The best results of additive model on the training samples are found at a=0.60 and b=0.40. Average Spearman correlation scores of both addition and multiplication models over the testing samples are displayed in table 5. The goodness of fit R 2 values when trained over the whole dataset are also displayed. Vector addition has a clear upper hand over multiplication in terms of both accuracy and goodness of fit for phrase compositionality prediction.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Composition function based models evaluation",
"sec_num": null
},
{
"text": "For phrase compositionality prediction (table 5), both constituent based and compositionality function based models are found to be competitive, though compositionality function based models perform slightly better. The reason could be because while constituent based models use contextual information of each constituent independently, composition function models make use of collective evidence from the contexts of both the constituents simultaneously. In the public evaluations of compositionality detection (Biemann and Giesbrecht, 2011) , our system (Reddy et al., 2011) which uses the notion of contexts salient to both the constituents achieved better performance than the system which uses only one of the constituent's contexts.",
"cite_spans": [
{
"start": 512,
"end": 542,
"text": "(Biemann and Giesbrecht, 2011)",
"ref_id": "BIBREF1"
},
{
"start": 556,
"end": 576,
"text": "(Reddy et al., 2011)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Winner",
"sec_num": null
},
{
"text": "All the results when compared with random baseline (RAND in table 5), which assigns a random compositionality score to a compound, are highly significant.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Winner",
"sec_num": null
},
{
"text": "In this paper we examined the compositionality judgments of noun compounds and also the literality judgments of their constituent words. Our study reveals that both the constituent words play a major role in deciding the compositionality of the phrase. We showed that the functions which predict the compositionality using both the constituent literality scores have high correlations with compositionality judgments. Based on this evidence we proposed constituent based models for compositionality detection. We compared constituent based models with compositionality function based models. The additive compositionality functions were slightly superior to the best performing constituent models (again additive) but performance is comparable and we plan to examine more sophisticated constituent models in the future.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "5"
},
{
"text": "All the 8100 annotations collected in this work are released publicly. We hope the dataset can reveal more insights into the compositionality in terms of the contribution from the constituents. Future directions of this work include token based disambiguation of phrases and designing more sophisticated constituent based models. Extending this study on other kinds of phrases such as adjective-noun, verb particle, verb-noun phrases may throw more light into our understanding of compositionality.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "5"
},
{
"text": "A study on AMT spammers http://bit.ly/ e1IPil2 The qualification test details are provided with the dataset. Please refer to footnote 3.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Annotation guidelines, Mechanical Turk hits, qualification test, annotators demographic and educational background, and final annotations are downloadable from http://sivareddy.in/downloads or http: //www.dianamccarthy.co.uk/downloads.html",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "This is similar to pointwise mutual information without logarithm",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Please note that simple additive model(Mitchell and Lapata, 2008) is different from the additive model described in(Guevara, 2011). In(Mitchell and Lapata, 2008) the coefficients are real numbers whereas in (Guevara, 2011) they are matrices.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "A statistical approach to the semantics of verb-particles",
"authors": [
{
"first": "Colin",
"middle": [],
"last": "Bannard",
"suffix": ""
},
{
"first": "Timothy",
"middle": [],
"last": "Baldwin",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Lascarides",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the ACL 2003 workshop on Multiword expressions: analysis, acquisition and treatment, MWE '03",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Colin Bannard, Timothy Baldwin, and Alex Las- carides. 2003. A statistical approach to the seman- tics of verb-particles. In Proceedings of the ACL 2003 workshop on Multiword expressions: analysis, acquisition and treatment, MWE '03.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Distributional Semantics and Compositionality 2011: Shared Task Description and Results",
"authors": [
{
"first": "Chris",
"middle": [],
"last": "Biemann",
"suffix": ""
},
{
"first": "Eugenie",
"middle": [],
"last": "Giesbrecht",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of DISCo-2011 in conjunction with ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chris Biemann and Eugenie Giesbrecht. 2011. Dis- tributional Semantics and Compositionality 2011: Shared Task Description and Results. In Proceed- ings of DISCo-2011 in conjunction with ACL 2011.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Pulling their weight: exploiting syntactic forms for the automatic identification of idiomatic expressions in context",
"authors": [
{
"first": "Paul",
"middle": [],
"last": "Cook",
"suffix": ""
},
{
"first": "Afsaneh",
"middle": [],
"last": "Fazly",
"suffix": ""
},
{
"first": "Suzanne",
"middle": [],
"last": "Stevenson",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the Workshop on a Broader Perspective on Multiword Expressions, MWE '07",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Paul Cook, Afsaneh Fazly, and Suzanne Stevenson. 2007. Pulling their weight: exploiting syntactic forms for the automatic identification of idiomatic expressions in context. In Proceedings of the Work- shop on a Broader Perspective on Multiword Ex- pressions, MWE '07.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Unsupervised type and token identification of idiomatic expressions",
"authors": [
{
"first": "Afsaneh",
"middle": [],
"last": "Fazly",
"suffix": ""
},
{
"first": "Paul",
"middle": [],
"last": "Cook",
"suffix": ""
},
{
"first": "Suzanne",
"middle": [],
"last": "Stevenson",
"suffix": ""
}
],
"year": 2009,
"venue": "Computional. Linguististics",
"volume": "35",
"issue": "",
"pages": "61--103",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Afsaneh Fazly, Paul Cook, and Suzanne Stevenson. 2009. Unsupervised type and token identification of idiomatic expressions. Computional. Linguististics., 35:61-103, March.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "WordNet: An Electronic Lexical Database",
"authors": [],
"year": 1998,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Christiane Fellbaum, editor. 1998. WordNet: An Elec- tronic Lexical Database. The MIT Press.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Introducing and evaluating ukWaC, a very large web-derived corpus of English",
"authors": [
{
"first": "A",
"middle": [],
"last": "Ferraresi",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Zanchetta",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Baroni",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Bernardini",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of the WAC4 Workshop at LREC",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Ferraresi, E. Zanchetta, M. Baroni, and S. Bernar- dini. 2008. Introducing and evaluating ukWaC, a very large web-derived corpus of English. In Pro- ceedings of the WAC4 Workshop at LREC 2008.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Understanding and Literal Meaning",
"authors": [
{
"first": "Raymond",
"middle": [
"W"
],
"last": "Gibbs",
"suffix": ""
}
],
"year": 1989,
"venue": "Cognitive Science",
"volume": "13",
"issue": "2",
"pages": "243--251",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Raymond W. Gibbs. 1989. Understanding and Literal Meaning. Cognitive Science, 13(2):243-251.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "In Search of Semantic Compositionality in Vector Spaces",
"authors": [
{
"first": "Eugenie",
"middle": [],
"last": "Giesbrecht",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the 17th International Conference on Conceptual Structures: Conceptual Structures: Leveraging Semantic Technologies, ICCS '09",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Eugenie Giesbrecht. 2009. In Search of Semantic Compositionality in Vector Spaces. In Proceedings of the 17th International Conference on Conceptual Structures: Conceptual Structures: Leveraging Se- mantic Technologies, ICCS '09.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Computing Semantic Compositionality in Distributional Semantics",
"authors": [
{
"first": "Raul",
"middle": [],
"last": "Emiliano",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Guevara",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the Ninth International Conference on Computational Semantics, IWCS",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Emiliano Raul Guevara. 2011. Computing Seman- tic Compositionality in Distributional Semantics. In Proceedings of the Ninth International Conference on Computational Semantics, IWCS '2011.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Automatic identification of non-compositional multiword expressions using latent semantic analysis",
"authors": [
{
"first": "Graham",
"middle": [],
"last": "Katz",
"suffix": ""
},
{
"first": "Eugenie",
"middle": [],
"last": "Giesbrecht",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the Workshop on Multiword Expressions: Identifying and Exploiting Underlying Properties, MWE '06",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Graham Katz and Eugenie Giesbrecht. 2006. Au- tomatic identification of non-compositional multi- word expressions using latent semantic analysis. In Proceedings of the Workshop on Multiword Expres- sions: Identifying and Exploiting Underlying Prop- erties, MWE '06.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Detecting compositionality in multi-word expressions",
"authors": [
{
"first": "Ioannis",
"middle": [],
"last": "Korkontzelos",
"suffix": ""
},
{
"first": "Suresh",
"middle": [],
"last": "Manandhar",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the ACL-IJCNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ioannis Korkontzelos and Suresh Manandhar. 2009. Detecting compositionality in multi-word expres- sions. In Proceedings of the ACL-IJCNLP 2009",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Automatic identification of noncompositional phrases",
"authors": [
{
"first": "Dekang",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 1999,
"venue": "Proceedings of the ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dekang Lin. 1999. Automatic identification of non- compositional phrases. In Proceedings of the ACL 1999.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Detecting a continuum of compositionality in phrasal verbs",
"authors": [
{
"first": "Diana",
"middle": [],
"last": "Mccarthy",
"suffix": ""
},
{
"first": "Bill",
"middle": [],
"last": "Keller",
"suffix": ""
},
{
"first": "John",
"middle": [],
"last": "Carroll",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the ACL 2003 workshop on Multiword expressions: analysis, acquisition and treatment, MWE '03",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Diana McCarthy, Bill Keller, and John Carroll. 2003. Detecting a continuum of compositionality in phrasal verbs. In Proceedings of the ACL 2003 workshop on Multiword expressions: analysis, ac- quisition and treatment, MWE '03.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Detecting Compositionality of Verb-Object Combinations using Selectional Preferences",
"authors": [
{
"first": "",
"middle": [],
"last": "Joshi",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of EMNLP-CoNLL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Joshi. 2007. Detecting Compositionality of Verb- Object Combinations using Selectional Preferences. In Proceedings of EMNLP-CoNLL 2007.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Vector-based Models of Semantic Composition",
"authors": [
{
"first": "Jeff",
"middle": [],
"last": "Mitchell",
"suffix": ""
},
{
"first": "Mirella",
"middle": [],
"last": "Lapata",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of ACL-08: HLT",
"volume": "",
"issue": "",
"pages": "236--244",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jeff Mitchell and Mirella Lapata. 2008. Vector-based Models of Semantic Composition. In Proceedings of ACL-08: HLT, pages 236-244, Columbus, Ohio.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Identifying Collocations to Measure Compositionality : Shared Task System Description",
"authors": [
{
"first": "Ted",
"middle": [],
"last": "Pedersen",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of DISCo-2011 in conjunction with ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ted Pedersen. 2011. Identifying Collocations to Mea- sure Compositionality : Shared Task System De- scription . In Proceedings of DISCo-2011 in con- junction with ACL 2011.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Exemplar-Based Word-Space Model for Compositionality Detection: Shared Task System Description",
"authors": [
{
"first": "Siva",
"middle": [],
"last": "Reddy",
"suffix": ""
},
{
"first": "Diana",
"middle": [],
"last": "Mccarthy",
"suffix": ""
},
{
"first": "Suresh",
"middle": [],
"last": "Manandhar",
"suffix": ""
},
{
"first": "Spandana",
"middle": [],
"last": "Gella",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the ACL Workshop on Distributional Semantics and Compositionality",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Siva Reddy, Diana McCarthy, Suresh Manandhar, and Spandana Gella. 2011. Exemplar-Based Word-Space Model for Compositionality Detection: Shared Task System Description. In Proceedings of the ACL Workshop on Distributional Semantics and Compositionality.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Multiword Expressions: A Pain in the Neck for NLP",
"authors": [
{
"first": "A",
"middle": [],
"last": "Ivan",
"suffix": ""
},
{
"first": "Timothy",
"middle": [],
"last": "Sag",
"suffix": ""
},
{
"first": "Francis",
"middle": [],
"last": "Baldwin",
"suffix": ""
},
{
"first": "Ann",
"middle": [
"A"
],
"last": "Bond",
"suffix": ""
},
{
"first": "Dan",
"middle": [],
"last": "Copestake",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Flickinger",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the CICLing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ivan A. Sag, Timothy Baldwin, Francis Bond, Ann A. Copestake, and Dan Flickinger. 2002. Multiword Expressions: A Pain in the Neck for NLP. In Pro- ceedings of the CICLing 2002.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "The Word-Space Model: Using distributional analysis to represent syntagmatic and paradigmatic relations between words in highdimensional vector spaces",
"authors": [
{
"first": "Magnus",
"middle": [],
"last": "Sahlgren",
"suffix": ""
}
],
"year": 2006,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Magnus Sahlgren. 2006. The Word-Space Model: Us- ing distributional analysis to represent syntagmatic and paradigmatic relations between words in high- dimensional vector spaces. Ph.D. thesis, Stockholm University.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Is Knowledge-Free Induction of Multiword Unit Dictionary Headwords a Solved Problem?",
"authors": [
{
"first": "Patrick",
"middle": [],
"last": "Schone",
"suffix": ""
},
{
"first": "Daniel",
"middle": [],
"last": "Jurafsky",
"suffix": ""
}
],
"year": 2001,
"venue": "Proceedings of EMNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Patrick Schone and Daniel Jurafsky. 2001. Is Knowledge-Free Induction of Multiword Unit Dic- tionary Headwords a Solved Problem? In Proceed- ings of EMNLP 2001.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Cheap and fast-but is it good?: evaluating non-expert annotations for natural language tasks",
"authors": [
{
"first": "Rion",
"middle": [],
"last": "Snow",
"suffix": ""
},
{
"first": "O'",
"middle": [],
"last": "Brendan",
"suffix": ""
},
{
"first": "Daniel",
"middle": [],
"last": "Connor",
"suffix": ""
},
{
"first": "Andrew",
"middle": [
"Y"
],
"last": "Jurafsky",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Ng",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of EMNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rion Snow, Brendan O'Connor, Daniel Jurafsky, and Andrew Y. Ng. 2008. Cheap and fast-but is it good?: evaluating non-expert annotations for natu- ral language tasks. In Proceedings of EMNLP 2008.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Unsupervised recognition of literal and non-literal use of idiomatic expressions",
"authors": [
{
"first": "Caroline",
"middle": [],
"last": "Sporleder",
"suffix": ""
},
{
"first": "Linlin",
"middle": [],
"last": "Li",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the EACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Caroline Sporleder and Linlin Li. 2009. Unsupervised recognition of literal and non-literal use of idiomatic expressions. In Proceedings of the EACL 2009.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "On the compositional and noncompositional nature of idiomatic expressions",
"authors": [
{
"first": "Debra",
"middle": [
"A"
],
"last": "Titone",
"suffix": ""
},
{
"first": "Cynthia",
"middle": [
"M"
],
"last": "Connine",
"suffix": ""
}
],
"year": 1999,
"venue": "Literal and Figurative Language",
"volume": "31",
"issue": "",
"pages": "1655--1674",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Debra A. Titone and Cynthia M. Connine. 1999. On the compositional and noncompositional nature of idiomatic expressions. Journal of Pragmatics, 31(12):1655 -1674. Literal and Figurative Lan- guage.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Measuring the relative compositionality of verbnoun (V-N) collocations by integrating features",
"authors": [
{
"first": "Sriram",
"middle": [],
"last": "Venkatapathy",
"suffix": ""
},
{
"first": "Aravind",
"middle": [
"K"
],
"last": "Joshi",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the HLT-EMNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sriram Venkatapathy and Aravind K. Joshi. 2005. Measuring the relative compositionality of verb- noun (V-N) collocations by integrating features. In Proceedings of the HLT-EMNLP 2005.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Semantic Vector Products: Some Initial Investigations",
"authors": [
{
"first": "Dominic",
"middle": [],
"last": "Widdows",
"suffix": ""
}
],
"year": 2008,
"venue": "Second AAAI Symposium on Quantum Interaction",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dominic Widdows. 2008. Semantic Vector Products: Some Initial Investigations. In Second AAAI Sympo- sium on Quantum Interaction, Oxford, March.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"text": "Mean values of phrase-level compositionality scores",
"uris": null,
"num": null
},
"TABREF1": {
"content": "<table><tr><td>Function f</td><td>\u03c1</td><td>R 2</td></tr><tr><td>ADD</td><td colspan=\"2\">0.966 0.937</td></tr><tr><td>MULT</td><td colspan=\"2\">0.965 0.904</td></tr><tr><td>COMB</td><td colspan=\"2\">0.971 0.955</td></tr><tr><td>WORD1</td><td colspan=\"2\">0.767 0.609</td></tr><tr><td>WORD2</td><td colspan=\"2\">0.720 0.508</td></tr></table>",
"num": null,
"type_str": "table",
"text": "Amazon Mechanical Turk statistics",
"html": null
},
"TABREF2": {
"content": "<table/>",
"num": null,
"type_str": "table",
"text": "Correlations between functions and phrase compositionality scores table 1. It is evident that annotators agree more at word level than phrase level annotations.",
"html": null
},
"TABREF3": {
"content": "<table><tr><td>Compound climate change 4.human being Word1 4.86\u00b10.34 4.33\u00b11.14 4.59\u00b10.72 interest rate Word2 Phrase Compound radio station 4.66\u00b10.96 4.34\u00b10.80 4.47\u00b10.72 health insurance law firm 4.72\u00b10.52 3.89\u00b11.50 4.40\u00b10.76 public service end user 3.87\u00b11.12 4.87\u00b10.34 4.25\u00b10.87 car park role model 3.55\u00b11.22 4.00\u00b11.03 4.11\u00b11.07 head teacher fashion plate 4.41\u00b11.07 3.31\u00b12.07 3.90\u00b11.42 balance sheet china clay 2.00\u00b11.84 4.62\u00b11.00 3.85\u00b11.27 game plan brick wall 3.16\u00b12.20 3.53\u00b11.86 3.79\u00b11.75 web site brass ring 3.73\u00b11.95 3.87\u00b11.98 3.72\u00b11.84 case study polo shirt 1.73\u00b11.41 5.00\u00b10.00 3.37\u00b11.38 rush hour search engine 4.62\u00b10.96 2.25\u00b11.70 3.32\u00b11.16 cocktail dress face value 1.39\u00b11.11 4.64\u00b10.81 3.04\u00b10.88 chain reaction cheat sheet 2.30\u00b11.59 4.00\u00b10.83 2.89\u00b11.11 blame game fine line 3.17\u00b11.34 2.03\u00b11.52 2.69\u00b11.21 front runner grandfather clock 0.43\u00b10.78 5.00\u00b10.00 2.64\u00b11.32 lotus position spelling bee 4.81\u00b10.77 0.52\u00b11.04 2.45\u00b11.25 silver screen smoking jacket 1.04\u00b10.82 4.90\u00b10.30 2.32\u00b11.29 spinning jenny number crunching 4.48\u00b10.77 0.97\u00b11.13 2.26\u00b11.00 guilt trip memory lane 4.75\u00b10.51 0.71\u00b10.80 2.17\u00b11.04 crash course rock bottom 0.74\u00b10.89 3.80\u00b11.08 2.14\u00b11.19 think tank night owl 4.47\u00b10.88 0.50\u00b10.82 1.93\u00b11.27 panda car diamond wedding 1.07\u00b11.29 3.41\u00b11.34 1.70\u00b11.05 firing line pecking order 0.78\u00b10.92 3.89\u00b11.40 1.69\u00b10.88 lip service cash cow 4.22\u00b11.07 0.37\u00b10.73 1.56\u00b11.10 graveyard shift sacred cow 1.93\u00b11.65 0.96\u00b11.72 1.52\u00b11.52 silver spoon flea market 0.38\u00b10.81 4.71\u00b10.84 1.52\u00b11.13 eye candy rocket science 0.64\u00b10.97 1.55\u00b11.40 1.43\u00b11.35 couch potato kangaroo court 0.17\u00b10.37 4.43\u00b11.02 1.37\u00b11.05 snail mail crocodile tears 0.19\u00b10.47 3.79\u00b11.05 1.25\u00b11.09 cutting edge zebra crossing 0.76\u00b10.62 4.61\u00b10.86 1.25\u00b11.02 acid test shrinking violet 2.28\u00b11.44 0.23\u00b10.56 1.07\u00b11.01 sitting duck rat race 0.25\u00b10.51 2.04\u00b11.32 0.86\u00b10.99 swan song gold mine 1.38\u00b11.42 0.70\u00b10.81 0.81\u00b10.82 rat run nest egg 0.79\u00b10.98 0.50\u00b10.87 0.78\u00b10.87 agony aunt snake oil 0.37\u00b10.55 0.81\u00b11.25 0.75\u00b11.12 monkey business 0.67\u00b11.01 1.85\u00b11.30 0.72\u00b10.69 Word1 Word2 Phrase 4.34\u00b10.99 4.69\u00b10.53 4.57\u00b10.90 4.53\u00b10.88 4.83\u00b10.58 4.40\u00b11.17 4.67\u00b10.65 4.77\u00b10.62 4.40\u00b10.76 4.90\u00b10.40 4.00\u00b11.10 4.20\u00b11.05 2.93\u00b11.51 4.52\u00b11.07 4.00\u00b11.16 3.82\u00b10.89 3.90\u00b10.96 3.86\u00b11.01 2.82\u00b11.96 4.86\u00b10.34 3.83\u00b11.23 2.68\u00b11.69 3.93\u00b11.18 3.79\u00b11.21 3.66\u00b11.12 4.67\u00b10.47 3.70\u00b10.97 3.11\u00b11.37 2.86\u00b11.36 3.33\u00b11.27 1.40\u00b11.08 5.00\u00b10.00 3.04\u00b11.22 2.41\u00b11.16 4.52\u00b10.72 2.93\u00b11.14 4.61\u00b10.67 2.00\u00b11.28 2.72\u00b10.92 3.97\u00b10.96 1.29\u00b11.10 2.66\u00b11.32 1.11\u00b11.17 4.78\u00b10.42 2.48\u00b11.22 1.41\u00b11.57 3.23\u00b11.45 2.38\u00b11.63 4.67\u00b10.54 0.41\u00b10.77 2.28\u00b11.08 4.71\u00b10.59 0.86\u00b10.94 2.19\u00b11.16 0.96\u00b10.94 4.23\u00b10.92 2.14\u00b11.27 3.96\u00b11.06 0.47\u00b10.62 2.04\u00b11.13 0.50\u00b10.56 4.66\u00b11.15 1.81\u00b11.07 1.61\u00b11.65 1.89\u00b11.50 1.70\u00b11.72 2.03\u00b11.25 1.75\u00b11.40 1.62\u00b11.06 0.38\u00b10.61 4.50\u00b10.72 1.52\u00b11.17 1.59\u00b11.47 1.44\u00b11.77 1.52\u00b11.45 3.83\u00b11.05 0.71\u00b10.75 1.48\u00b11.10 3.27\u00b11.48 0.34\u00b10.66 1.41\u00b11.03 0.60\u00b10.80 4.59\u00b11.10 1.31\u00b11.02 0.88\u00b11.19 1.73\u00b11.63 1.25\u00b11.18 0.71\u00b11.10 3.90\u00b11.24 1.22\u00b11.26 1.48\u00b11.48 0.41\u00b10.67 0.96\u00b11.04 0.38\u00b10.61 1.11\u00b11.14 0.83\u00b10.91 0.41\u00b10.62 2.33\u00b11.40 0.79\u00b10.66 1.86\u00b11.22 0.43\u00b10.56 0.76\u00b10.86 smoking gun 0.71\u00b10.75 1.00\u00b10.94 0.71\u00b10.84 silver bullet 0.52\u00b11.00 0.55\u00b11.10 0.67\u00b11.15 melting pot 1.00\u00b11.15 0.48\u00b10.63 0.54\u00b10.63 ivory tower 0.38\u00b11.03 0.54\u00b10.68 0.46\u00b10.68 cloud nine 0.47\u00b10.62 0.23\u00b10.42 0.33\u00b10.54 gravy train 0.30\u00b10.46 0.45\u00b10.77 0.31\u00b10.59</td></tr></table>",
"num": null,
"type_str": "table",
"text": "90\u00b10.30 4.83\u00b10.38 4.97\u00b10.18 engine room 4.86\u00b10.34 5.00\u00b10.00 4.93\u00b10.25 graduate student 4.70\u00b10.46 5.00\u00b10.00 4.90\u00b10.30 swimming pool 4.80\u00b10.40 4.90\u00b10.30 4.87\u00b10.34 speed limit 4.93\u00b10.25 4.83\u00b10.38 4.83\u00b10.46 research project 4.90\u00b10.30 4.53\u00b10.96 4.82\u00b10.38 application form 4.77\u00b10.42 4.86\u00b10.34 4.80\u00b10.48 bank account 4.87\u00b10.34 4.83\u00b10.46 4.73\u00b10.44 parking lot 4.83\u00b10.37 4.77\u00b10.50 4.70\u00b10.64 credit card 4.67\u00b10.54 4.90\u00b10.30 4.67\u00b10.70 ground floor 4.66\u00b10.66 4.70\u00b10.78 4.67\u00b10.60 mailing list 4.67\u00b10.54 4.93\u00b10.25 4.67\u00b10.47 call centre 4.73\u00b10.44 4.41\u00b10.72 4.66\u00b10.66 video game 4.50\u00b10.72 5.00\u00b10.00 4.60\u00b10.61",
"html": null
},
"TABREF4": {
"content": "<table/>",
"num": null,
"type_str": "table",
"text": "Compounds with their constituent and phrase level mean\u00b1deviation scores one testing sample at each iteration). The coefficients of the functions are estimated using leastsquare linear regression technique over the training samples. The average Spearman correlation scores (\u03c1) over testing samples are displayed in table 3. The goodness of fit R 2 values when trained over the whole data are also displayed in table 3.",
"html": null
},
"TABREF6": {
"content": "<table><tr><td>: Phrase level correlations of composition-</td></tr><tr><td>ality scores</td></tr><tr><td>fit values. The model MULT shows good correla-</td></tr><tr><td>tion but the goodness of fit is lower. First con-</td></tr><tr><td>stituent (model WORD1 i.e.</td></tr></table>",
"num": null,
"type_str": "table",
"text": "",
"html": null
}
}
}
} |