File size: 106,090 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 |
{
"paper_id": "N15-1006",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T14:34:38.466002Z"
},
"title": "An Incremental Algorithm for Transition-based CCG Parsing",
"authors": [
{
"first": "Bharat",
"middle": [
"Ram"
],
"last": "Ambati",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Edinburgh",
"location": {}
},
"email": "bharat.ambati@ed.ac.uk"
},
{
"first": "Tejaswini",
"middle": [],
"last": "Deoskar",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Edinburgh",
"location": {}
},
"email": "tdeoskar@inf.ed.ac.uk"
},
{
"first": "Mark",
"middle": [],
"last": "Johnson",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Macquarie University",
"location": {}
},
"email": "mark.johnson@mq.edu.au"
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Edinburgh",
"location": {}
},
"email": "steedman@inf.ed.ac.uk"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Incremental parsers have potential advantages for applications like language modeling for machine translation and speech recognition. We describe a new algorithm for incremental transition-based Combinatory Categorial Grammar parsing. As English CCGbank derivations are mostly right branching and non-incremental, we design our algorithm based on the dependencies resolved rather than the derivation. We introduce two new actions in the shift-reduce paradigm based on the idea of 'revealing' (Pareschi and Steedman, 1987) the required information during parsing. On the standard CCGbank test data, our algorithm achieved improvements of 0.88% in labeled and 2.0% in unlabeled F-score over a greedy non-incremental shift-reduce parser.",
"pdf_parse": {
"paper_id": "N15-1006",
"_pdf_hash": "",
"abstract": [
{
"text": "Incremental parsers have potential advantages for applications like language modeling for machine translation and speech recognition. We describe a new algorithm for incremental transition-based Combinatory Categorial Grammar parsing. As English CCGbank derivations are mostly right branching and non-incremental, we design our algorithm based on the dependencies resolved rather than the derivation. We introduce two new actions in the shift-reduce paradigm based on the idea of 'revealing' (Pareschi and Steedman, 1987) the required information during parsing. On the standard CCGbank test data, our algorithm achieved improvements of 0.88% in labeled and 2.0% in unlabeled F-score over a greedy non-incremental shift-reduce parser.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Combinatory Categorial Grammar (CCG) (Steedman, 2000) is an efficiently parseable, yet linguistically expressive grammar formalism.",
"cite_spans": [
{
"start": 37,
"end": 53,
"text": "(Steedman, 2000)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In addition to predicate-argument structure, CCG elegantly captures the unbounded dependencies found in grammatical constructions like relativization, coordination etc. Availability of the English CCGbank (Hockenmaier and Steedman, 2007) has enabled the creation of several robust and accurate wide-coverage CCG parsers Clark and Curran, 2007; Zhang and Clark, 2011) . While the majority of CCG parsers use chart-based approaches Clark and Curran, 2007) , there has been some work on developing shift-reduce parsers for CCG (Zhang and Clark, 2011; Xu et al., 2014) . Most of these parsers model normal-form CCG derivations (Eisner, 1996) , which are mostly right-branching trees : hence are not incremental in nature. The dependency models of Clark and Curran (2007) and Xu et al. (2014) model dependencies rather than derivations, but do not guarantee incremental analyses.",
"cite_spans": [
{
"start": 205,
"end": 237,
"text": "(Hockenmaier and Steedman, 2007)",
"ref_id": "BIBREF13"
},
{
"start": 320,
"end": 343,
"text": "Clark and Curran, 2007;",
"ref_id": "BIBREF4"
},
{
"start": 344,
"end": 366,
"text": "Zhang and Clark, 2011)",
"ref_id": "BIBREF28"
},
{
"start": 430,
"end": 453,
"text": "Clark and Curran, 2007)",
"ref_id": "BIBREF4"
},
{
"start": 524,
"end": 547,
"text": "(Zhang and Clark, 2011;",
"ref_id": "BIBREF28"
},
{
"start": 548,
"end": 564,
"text": "Xu et al., 2014)",
"ref_id": "BIBREF26"
},
{
"start": 623,
"end": 637,
"text": "(Eisner, 1996)",
"ref_id": "BIBREF9"
},
{
"start": 743,
"end": 766,
"text": "Clark and Curran (2007)",
"ref_id": "BIBREF4"
},
{
"start": 771,
"end": 787,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Besides being cognitively plausible (Marslen-Wilson, 1973) , incremental parsing is more useful than non-incremental parsing for some applications. For example, an incremental analysis is required for integrating syntactic and semantic information into language modeling for statistical machine translation (SMT) and automatic speech recognition (ASR) (Roark, 2001 ; Wang and Harper, 2003) . This paper develops a new incremental shiftreduce algorithm for parsing CCG by building a dependency graph in addition to the CCG derivation as a representation. The dependencies in the graph are extracted from the CCG derivation. A node can have multiple parents, and hence we construct a dependency graph rather than a tree. Two new actions are introduced in the shift-reduce paradigm for \"revealing\" (Pareschi and Steedman, 1987) unbuilt structure during parsing. We build the dependency graph in parallel to the incremental CCG derivation and use this graph for revealing, via these two new actions. On the standard CCGbank test data, our algorithm achieves improvements of 0.88% in labeled F-score and 2.0% in unlabeled F-score over a greedy non-incremental shift-reduce algorithm. As our algorithm does not model derivations, but rather models transitions, we do not need a treebank John likes mangoes from India madly of incremental CCG derivations and can train on the dependencies in the existing treebank. Our approach can therefore be adapted to other languages with dependency treebanks, since CCG lexical categories can be easily extracted from dependency treebanks (Cakici, 2005; Ambati et al., 2013) . The rest of the paper is arranged as follows. Section 2 gives a brief introduction to related work in the areas of CCG parsing and incremental parsing. In section 3, we describe our incremental shift-reduce parsing algorithm. Details about the experiments, evaluation metrices and analysis of the results are in section 4. We conclude with possible future directions in section 5.",
"cite_spans": [
{
"start": 36,
"end": 58,
"text": "(Marslen-Wilson, 1973)",
"ref_id": "BIBREF17"
},
{
"start": 352,
"end": 364,
"text": "(Roark, 2001",
"ref_id": "BIBREF20"
},
{
"start": 367,
"end": 389,
"text": "Wang and Harper, 2003)",
"ref_id": "BIBREF25"
},
{
"start": 795,
"end": 824,
"text": "(Pareschi and Steedman, 1987)",
"ref_id": "BIBREF19"
},
{
"start": 1571,
"end": 1585,
"text": "(Cakici, 2005;",
"ref_id": "BIBREF2"
},
{
"start": 1586,
"end": 1606,
"text": "Ambati et al., 2013)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this section, we first give a brief introduction to various available CCG parsers. Then we describe approaches towards incremental and greedy parsing.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "There has been a significant amount of work on developing chart-based parsers for CCG. Both generative and discriminative (Clark et al., 2002; Clark and Curran, 2007; Auli and Lopez, 2011; Lewis and Steedman, 2014) models have been developed. As these parsers employ a bottom-up chart-parsing strategy and use normal-form CCGbank derivations which are rightbranching, they are not incremental in nature. In an SVO (Subject-Verb-Object) language, these parsers first attach the object to the verb and then the subject.",
"cite_spans": [
{
"start": 122,
"end": 142,
"text": "(Clark et al., 2002;",
"ref_id": "BIBREF5"
},
{
"start": 143,
"end": 166,
"text": "Clark and Curran, 2007;",
"ref_id": "BIBREF4"
},
{
"start": 167,
"end": 188,
"text": "Auli and Lopez, 2011;",
"ref_id": "BIBREF1"
},
{
"start": 189,
"end": 214,
"text": "Lewis and Steedman, 2014)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "CCG Parsers",
"sec_num": "2.1"
},
{
"text": "Two major works in shift-reduce CCG parsing with accuracies competitive with the widely used Clark and Curran (2007) parser (C&C) are Zhang and Clark (2011) and Xu et al. (2014) . Zhang and Clark (2011) used a global linear model trained discriminatively with the averaged perceptron (Collins, 2002) and beam search for their shiftreduce CCG parser. Xu et al. (2014) developed a dependency model for shift-reduce CCG parsing using a dynamic oracle technique. Unlike the chart parsers, both these parsers can produce fragmentary analyses when a complete spanning analysis is not found. Both these shift-reduce parsers are more incremental than standard chart based parsers. But, as they employ an arc-standard (Yamada and Matsumoto, 2003) shift-reduce strategy on CCGbank, given an SVO language, these parsers are not guaranteed to attach the subject before the object.",
"cite_spans": [
{
"start": 93,
"end": 116,
"text": "Clark and Curran (2007)",
"ref_id": "BIBREF4"
},
{
"start": 134,
"end": 156,
"text": "Zhang and Clark (2011)",
"ref_id": "BIBREF28"
},
{
"start": 161,
"end": 177,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
},
{
"start": 180,
"end": 202,
"text": "Zhang and Clark (2011)",
"ref_id": "BIBREF28"
},
{
"start": 284,
"end": 299,
"text": "(Collins, 2002)",
"ref_id": "BIBREF7"
},
{
"start": 350,
"end": 366,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
},
{
"start": 709,
"end": 737,
"text": "(Yamada and Matsumoto, 2003)",
"ref_id": "BIBREF27"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "CCG Parsers",
"sec_num": "2.1"
},
{
"text": "A strictly incremental parser is one which computes the relationship between words as soon as they are encountered in the input. Shift-reduce CCG parsers rely either on CCGbank derivations (Zhang and Clark, 2011) which are non-incremental, or on dependencies (Xu et al., 2014) which could be incremental in simple cases, but do not guarantee incrementality. Hassan et al. (2009) developed a semi-incremental CCG parser by transforming the English CCGbank into left branching derivation trees. The strictly incremental version performed with very low accuracy but a semi-incremental version gave a balance between incrementality and accuracy. There is also some work on incremental parsing using grammar formalisms other than CCG like phrase structure grammar (Collins and Roark, 2004) and tree substitution grammar (Sangati and Keller, 2013) .",
"cite_spans": [
{
"start": 189,
"end": 212,
"text": "(Zhang and Clark, 2011)",
"ref_id": "BIBREF28"
},
{
"start": 259,
"end": 276,
"text": "(Xu et al., 2014)",
"ref_id": "BIBREF26"
},
{
"start": 358,
"end": 378,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
},
{
"start": 759,
"end": 784,
"text": "(Collins and Roark, 2004)",
"ref_id": "BIBREF6"
},
{
"start": 815,
"end": 841,
"text": "(Sangati and Keller, 2013)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Incremental Parsers",
"sec_num": "2.2"
},
{
"text": "There has been a significant amount of work on greedy shift-reduce dependency parsing. The Malt parser (Nivre et al., 2007) is one of the earliest parsers based on this paradigm. Goldberg and Nivre (2012) improved learning for greedy parsers by using dynamic oracles rather than a single static transition sequence as the oracle. In all the standard shift-reduce parsers, when two trees combine, only the top node (root) of each tree participates in the action. Sartorio et al. (2013) introduced a technique where in addition to the root node, nodes on the right and left periphery respectively are also available for attachment in the parsing process. A non-monotonic parsing strategy was introduced by Honnibal et al. (2013) , where an action taken during the parsing process is revised based on future context. Figure 2 : NonInc -Sequence of actions with parser configuration and the corresponding dependency graph.",
"cite_spans": [
{
"start": 103,
"end": 123,
"text": "(Nivre et al., 2007)",
"ref_id": "BIBREF18"
},
{
"start": 179,
"end": 204,
"text": "Goldberg and Nivre (2012)",
"ref_id": "BIBREF10"
},
{
"start": 462,
"end": 484,
"text": "Sartorio et al. (2013)",
"ref_id": "BIBREF22"
},
{
"start": 704,
"end": 726,
"text": "Honnibal et al. (2013)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [
{
"start": 814,
"end": 822,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Greedy Parsers",
"sec_num": "2.3"
},
{
"text": "Though the performance of these greedy parsers is less accurate than related parsers using a beam (Zhang and Nivre, 2011) , greedy parsers are interesting as they are very fast and are practically useful in large-scale applications such as parsing the web and online machine translation or speech recognition. In this work, we develop a new greedy transition-based algorithm for incremental CCG parsing, which is more incremental than Zhang and Clark (2011) and Xu et al. (2014) and more accurate than Hassan et al. (2009) . Our algorithm is not strictly incremental as we only produce derivations which are compatible with the Strict Competence Hypothesis (Steedman, 2000) ",
"cite_spans": [
{
"start": 98,
"end": 121,
"text": "(Zhang and Nivre, 2011)",
"ref_id": "BIBREF29"
},
{
"start": 435,
"end": 457,
"text": "Zhang and Clark (2011)",
"ref_id": "BIBREF28"
},
{
"start": 462,
"end": 478,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
},
{
"start": 502,
"end": 522,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
},
{
"start": 657,
"end": 673,
"text": "(Steedman, 2000)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Greedy Parsers",
"sec_num": "2.3"
},
{
"text": "(details in \u00a73.2.3).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Greedy Parsers",
"sec_num": "2.3"
},
{
"text": "We first describe the Zhang and Clark (2011) style shift-reduce algorithm for CCG parsing. Then we explain our incremental algorithm based on the \"revealing\" technique for shift-reduce CCG parsing.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Algorithms",
"sec_num": "3"
},
{
"text": "This is our baseline algorithm and is similar to Zhang and Clark (2011)'s algorithm (henceforth NonInc). It consists of an input buffer and a stack and has four major parsing actions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "\u2022 Shift -X (S) : Pushes a word from the input buffer to the stack and assigns a CCG category X. This action performs category disambiguation as well, as X can be any of the categories assigned by a supertagger. \u2022 Reduce Left -X (RL) : Pops the top two nodes from the stack, combines them into a new node and pushes it back onto the stack with a category X. This corresponds to binary rules in the CCGbank (e.g. CCG combinators like function application, composition etc., and punctuation rules). In this action the right node is the head and hence the left node is reduced.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "\u2022 Reduce Right -X (RR) : This action is similar to the RL (Reduce Left -X) action, except that in this action the right node is reduced since the left node is the head.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "\u2022 Unary -X (U) : Pops the top node from the stack, converts it into a new node with category X and pushes it back on the stack. The head remains the same in this action. This action corresponds to unary rules in the CCGbank (unary type-changing and type-raising rules). Figure 1 shows a normal-form CCG derivation for an example sentence 'John likes mangoes from India madly'. Figure 2 shows the sequence of steps using the NonInc algorithm for parsing the sentence. For simplicity and space reasons, unary productions leading to NP are not described. From step 1 through step 5, the first five words in the sentence (John, likes, mangoes, from, India) are shifted with corresponding categories using shift actions (S). In step 6, (NP\\NP)/NP:from and NP:India are combined using the Reduce-Right (RR) action to form NP\\NP:from which is combined with NP:mangoes in step 7 to form NP:mangoes.",
"cite_spans": [],
"ref_spans": [
{
"start": 270,
"end": 278,
"text": "Figure 1",
"ref_id": "FIGREF0"
},
{
"start": 377,
"end": 385,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "Step 8 combines (S\\NP)/NP:likes with NP:mangoes to form S\\NP:likes using RR action. Then the next word 'madly' is shifted in step 9, which is then combined with S\\NP:likes in step 10. In step 11, NP:John and S\\NP:likes are combined using Reduce-Left (RL) action leading to S:likes. The parsing process terminates at this step as there are no more tokens in the input buffer and as there is only a single node left in the stack.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "(1) We use indexed CCG categories (Clark et al., 2002) and obtain the CCG dependencies after every action to build the dependency graph in parallel to the CCG derivation. This is similar to Xu et al. (2014) but differs from Zhang and Clark (2011) , who extract the dependencies at the end after obtaining a derivation for the entire sentence. Figure 2 also shows the dependency graph generated and the arc labels give the step ID after which the dependency is generated.",
"cite_spans": [
{
"start": 34,
"end": 54,
"text": "(Clark et al., 2002)",
"ref_id": "BIBREF5"
},
{
"start": 190,
"end": 206,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
},
{
"start": 224,
"end": 246,
"text": "Zhang and Clark (2011)",
"ref_id": "BIBREF28"
}
],
"ref_spans": [
{
"start": 343,
"end": 352,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "S [ NP John (2) S [ NP John (S\\NP)/NP likes (3) RL [ S/NP likes (4) S [ S/NP likes NPmangoes (5) RR [ S likes (6) S [ S likes (NP\\NP)/NP f rom (7) S [ S likes (NP\\NP)/NP f rom NP India (8) RR [ S likes NP\\NP f rom (9) RRev [ S likes (10) S [ S likes (S\\NP)\\(S\\NP) madly (11) LRev [ S likes",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Non Incremental Algorithm (NonInc)",
"sec_num": "3.1"
},
{
"text": "(RevInc)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing based Incremental Algorithm",
"sec_num": "3.2"
},
{
"text": "The NonInc algorithm described above is not incremental because it relies purely on the mostly rightbranching CCG derivation. In our example sentence, the verb (likes) combines with the subject (John) only at the end (step ID = 11) after all the remaining words in the sentence are processed, making the parse non-incremental. In this section we describe a new incremental algorithm based on a 'revealing' technique (Pareschi and Steedman, 1987) which tries to build the most incremental derivation.",
"cite_spans": [
{
"start": 416,
"end": 445,
"text": "(Pareschi and Steedman, 1987)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing based Incremental Algorithm",
"sec_num": "3.2"
},
{
"text": "Pareschi and Steedman (1987)'s original version of revealing was defined in terms of (implicitly higher-order) unification. It was based on the following observation. If we think of categories as terms in a logic programming language, then while we usually think of CCG combinatory rules like the following as applying with the two categories on the left X/Y and Y as inputs, say instantiated as S /NP and NP , to define the category X on the right as S, in fact instantiating any two of those categories defines the third.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "X/Y Y =\u21d2 X",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "For example, if we define X and X/Y as S and S /NP , we clearly define Y as NP . They proposed to use unification-based revealing to recover unbuilt constituents in from the result of overlygreedy incremental parsing. A related secondorder matching-based mechanism was used by (Kwiatkowski et al., 2010) to decompose logical forms for semantic parser induction.",
"cite_spans": [
{
"start": 277,
"end": 303,
"text": "(Kwiatkowski et al., 2010)",
"ref_id": "BIBREF15"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "The present incremental parser uses a related revealing technique confined to the right periphery. Using CCG combinators and rules like type-raising followed by forward composition, we combine nodes in the stack if there is a dependency between them. However, this can create problems for the newly shifted node as its dependent might already have been reduced. For instance, if the object 'mangoes' is reduced after it is shifted to the stack, then it won't be available for the preposition phrase (PP) 'from India' (of course, this goes for more complex NPs as well). We have to extract 'mangoes', which is hidden in the derivation, so as to make the correct attachment to the PP. This is where revealing comes into play. Mangoes is 'revealed' so that it is available to attach to the PP following it, although it has already been reduced. To handle this, in addition to the four actions of the NonInc algorithm, we introduce two new actions: Left Reveal (LRev) and Right Reveal (RRev). For this, after every action, in addition to updating the stack we also keep track of the dependencies resolved and update the dependency graph accordingly 1 . In other words, we build the dependency graph for the sentence in parallel to the CCG derivation. As these dependencies are extracted from the CCG derivation, a node can have multiple parents and hence we construct a dependency graph rather than a tree.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "S likes NP\\NP f rom R > S/NP likes NPmangoes < NP > S (a) RRev S likes (S\\NP)\\(S\\NP) madly R < NP John S\\NP likes < S\\NP < S (b) LRev",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "\u2022 Left Reveal (LRev) : Pop the top two nodes in the stack (left, right). Identify the left node's child with a subject dependency. Abstract over this child node and split the category of left node into two categories. Combine the nodes using CCG combinators accordingly. VP modifiers like VP coordination require this action.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "\u2022 Right Reveal (RRev) : Pop the top two nodes in the stack (left, right) . Check the right periphery of the left node in the dependency graph, extract all the nodes with compatible CCG categories and identify all the possible nodes that the right node can combine with. Abstract over this node (e.g. object), split the category into two categories accordingly and combine the nodes using CCG combinators. Constructions like NP coordination, and PP attachment require this action. Figure 3 shows the sequence of steps for the example sentence described above. In steps 1 and 2, the first two words in the sentence: 'John' and 'likes', are shifted from the input buffer to the stack. In addition to standard CCG combinators of application and composition, we also use type-raising followed by forward composition 2 . In step 3, the category of the left node 'John', NP, is type-raised to S/(S\\NP) which is then combined with the category of right node 'likes', (S\\NP)/NP, using forward composition operator to yield the category S/NP. This step also updates the dependency graph with an edge between 'John' and 'likes', where 'likes' is the parent and 'John' is the child. The next word 'mangoes' is shifted in step 4 and combined with S/NP:likes in step 5 using RR action yielding S:likes. After this step, the dependency graph will have 'likes' as the root, with 'John' and 'mangoes' as its children. In this way, as our algorithm tries to be more incremental, both subject and object arguments are resolved as soon as the corresponding tokens are shifted to the stack.",
"cite_spans": [
{
"start": 59,
"end": 72,
"text": "(left, right)",
"ref_id": null
}
],
"ref_spans": [
{
"start": 480,
"end": 488,
"text": "Figure 3",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Revealing",
"sec_num": "3.2.1"
},
{
"text": "In steps 6 and 7, the next two words 'from' and 'India' are shifted to the stack.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Worked Example",
"sec_num": "3.2.2"
},
{
"text": "Step 8 combines (NP\\NP)/NP:from and NP:India using RR action to form NP\\NP:from. Now, we apply the RRev action in step 9 to correctly attach 'from' to 'mangoes'. In RRev we first check the right periphery and identify a possible node to be attached, 'mangoes', which is the object argument of the verb 'likes'. We abstract over this object and split the category in the following manner: If X is the category of the left node and Y\\Y is the category of the right node, then X is split into X/Y and Y with corresponding heads. The head of the left node will be the head of X/Y, and the dependency graph helps in identifying the correct head for Y. Now, Y and Y\\Y can be combined using the backward application rule to form Y, which can be combined with X/Y to form X back. In our example sentence, S:likes is split into S/NP:likes and NP:mangoes. NP:mangoes is combined with NP\\NP:from to form NP:mangoes, which in return combines with S/NP:likes and forms back S:likes. Figure 4 (a) sketches this process. This action also updates the dependency graph with a dependency between 'mangoes' and 'from'. The next word 'madly' is shifted in step 10, after which the stack has two nodes S:likes and (S\\NP)\\(S\\NP):madly. We apply the LRev action to combine these two nodes. We abstract over the subject of the left node, 'likes', and split the category. Here, S:likes is split into NP:John and S\\NP:likes. S\\NP:likes is combined with (S\\NP)\\(S\\NP):madly to form S\\NP:likes, which in return combines with NP:John and forms back S:likes. The dependency graph is updated with a dependency between 'likes' and 'madly'. Note that the final output is a standard CCG tree. Figure 4 (b) shows this LRev action.",
"cite_spans": [],
"ref_spans": [
{
"start": 970,
"end": 978,
"text": "Figure 4",
"ref_id": "FIGREF2"
},
{
"start": 1659,
"end": 1667,
"text": "Figure 4",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Worked Example",
"sec_num": "3.2.2"
},
{
"text": "Our incremental algorithm uses a combination of the CCG derivation and a dependency graph that helps to 'reveal' unbuilt structure in the CCG derivation by identifying heads of the revealed categories. For example in Figure-4a , in RRev action, S:likes is split into S/NP:likes and NP:mangoes. The splitting of categories is deterministic but the right periphery of the dependency graph helps in identifying the head, which is 'mangoes'. The theoretical idea of 'revealing' is from Pareschi and Steedman (1987) , but they used only a toy grammar without a model or empirical results. Checking the right periphery is similar to Sartorio et al. (2013) and abstracting over the left or right argument is similar to Dalrymple et al. (1991) . Currently, we abstract only over arguments. Adding a new action to abstract over the verb as well will make our algorithm handle ellipses in the sentences like 'John likes mangoes and Mary too' similar to Dalrymple et al. (1991) but we leave that for future work.",
"cite_spans": [
{
"start": 482,
"end": 510,
"text": "Pareschi and Steedman (1987)",
"ref_id": "BIBREF19"
},
{
"start": 627,
"end": 649,
"text": "Sartorio et al. (2013)",
"ref_id": "BIBREF22"
},
{
"start": 712,
"end": 735,
"text": "Dalrymple et al. (1991)",
"ref_id": "BIBREF8"
},
{
"start": 943,
"end": 966,
"text": "Dalrymple et al. (1991)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [
{
"start": 217,
"end": 226,
"text": "Figure-4a",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Analysis",
"sec_num": "3.2.3"
},
{
"text": "Our system is monotonic in the sense that the set of dependency relationships grows monotonically during the parsing process. Our algorithm gives derivations almost as incremental as Hassan et al. (2009) but without changing the lexical categories and without backtracking. The only change we made to the CCGbank is making the main verb the head of the auxiliary rather than the reverse as in CCGbank derivations. In the right derivational trees of CCGbank, the main verb is the head for its right side arguments and the auxiliary verb is the head for the left side arguments in the derivation. Not changing the head rule would make our algorithm use the costly reveal actions significantly more, which we avoid by changing the head direction. 3% of the total dependencies are affected by this modification.",
"cite_spans": [
{
"start": 183,
"end": 203,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Analysis",
"sec_num": "3.2.3"
},
{
"text": "Though our algorithm can be completely incremental, we currently compromise incrementality in the following cases: (a) no dependency between the nodes in the stack (b) unary type-changing and non-standard binary rules (c) adjuncts like VP modifiers and coordinate constructions like VP, sentential coordination.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Analysis",
"sec_num": "3.2.3"
},
{
"text": "We find empirically that extending incrementality to cover these cases actually reduces parsing performance significantly. It also violates the Strict Competence Hypothesis (SCH) (Steedman, 2000) , which argues on evolutionary and developmental grounds that the parser can only build constituents that are typable by the competence grammar. We explored the adjunct case of attaching only the preposition first rather than creating a complete prepositional phrase and then attaching it to correct parent. In our example sentence, this would be the case of attaching the preposition 'from' to its parent using RRev and then combining the NP 'India' accordingly as opposed to creating the preposition phrase 'from India' first and then using RRev action to attach it to the correct parent. Though the former is more incremental, it is inconsistent with the SCH. The latter analysis is consistent with strict competence and also gave better parsing performance while compromising incrementality only slightly. The empirical impact of these differing degrees of incrementality on extrinsic evaluation of our algorithm in terms of language modeling for SMT or ASR is left for future work.",
"cite_spans": [
{
"start": 179,
"end": 195,
"text": "(Steedman, 2000)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Analysis",
"sec_num": "3.2.3"
},
{
"text": "Using our incremental algorithm, we converted the CCGbank derivations into a sequence of shiftreduce actions. We could convert around 98% of the derivations, which is the coverage of our algorithm, recovering around 99% dependencies. Problematic cases are mainly the ones which involve nonstandard binary rules, and punctuations with lexical CCG categories other than 'conj', used as a conjunction, or ',' which is treated as a punctuation mark.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Analysis",
"sec_num": "3.2.3"
},
{
"text": "We re-implemented Zhang and Clark (2011)'s model for our experiments. We used their global linear model trained with the averaged perceptron (Collins, 2002) . We applied the early-update strategy of Collins and Roark (2004) while training. In this strategy, when we don't use a beam, decoding is stopped when the predicted action is different from the gold action and weights are updated accordingly. We use the feature set of Zhang and Clark (2011) (Z&C) for the NonInc algorithm. This feature set comprises of features over the top four nodes in the stack and the next four words in the input buffer. Complete details of the feature set can be found in their paper. For our own model, RevInc, in addition to these features used for NonInc, we also provide features based on the right periphery of top node in the stack. For nodes in the right periphery, we provide uni-gram and bi-gram features based on the node's CCG category. For example, if S0 is the node on the top of the stack, B1 is the bottom most node in the right periphery, and c represent the node's CCG category, then B1c, and B1cS0c are the uni-gram and bi-gram features respectively.",
"cite_spans": [
{
"start": 141,
"end": 156,
"text": "(Collins, 2002)",
"ref_id": "BIBREF7"
},
{
"start": 199,
"end": 223,
"text": "Collins and Roark (2004)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Results",
"sec_num": "4"
},
{
"text": "Unlike Z&C, we do not use a beam for our experiments, although we use a beam of 16 for comparison of our results with their parser. The latter gives competitive results with the state-of-theart CCG parsers. Z&C and Xu et al. (2014) , use C&C's generate script and unification mechanism respectively to extract dependencies for evaluation. C&C's grammar doesn't cover all the lexical categories and binary rules in the CCGbank. To avoid this, we adapted Hockenmaier's scripts used for extracting dependencies from the CCGbank derivations. These are the two major divergences in our re-implementation from Z&C.",
"cite_spans": [
{
"start": 215,
"end": 231,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Results",
"sec_num": "4"
},
{
"text": "We use standard CCGbank training (sections 02 \u2212 21), development (section 00) and testing (section 23) splits for our experiments. All sentences in the training set are used to train NonInc. But for RevInc, we used 98% of the training set (the coverage of our algorithm). We use automatic POS-tags and lexical CCG categories assigned using the C&C POS tagger and supertagger respectively for development and test data. For training data, these tags are assigned using ten-way jackknifing. Also, for lexical CCG categories, we use a multitagger which assigns k-best supertags to a word rather than 1-best supertagging (Clark and Curran, 2004) . The number of supertags assigned to a word depends on a \u03b2 parameter. Unlike Z&C, the default value of \u03b2 gave us better results rather than decreasing the value. Not using a beam could be the reason for this.",
"cite_spans": [
{
"start": 617,
"end": 641,
"text": "(Clark and Curran, 2004)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data and Settings",
"sec_num": "4.1"
},
{
"text": "Following Z&C and Xu et al. (2014) , during training, we also provide the gold CCG lexical category to the list of CCG lexical categories for a word if it is not assigned by the supertagger.",
"cite_spans": [
{
"start": 10,
"end": 34,
"text": "Z&C and Xu et al. (2014)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data and Settings",
"sec_num": "4.1"
},
{
"text": "Before evaluating the performance of our algorithm, we introduce two measures of incrementality: connectedness and waiting time. In a shift-reduce parser, a derivation is fully connected when all the nodes in the stack are connected leading to only one node in the stack at any point of time. We measure the average number of nodes in the stack before shifting a new token from input buffer to the stack, which we call as connectedness. For a fully connected incremental parser like Hassan et al. (2009) , connectedness would be one. As our RevInc algorithm is not fully connected, this number will be greater than one. For example, in a noun phrase 'the big book', when 'the' and 'big' are in the stack, as there is no dependency between these two words, our algorithm doesn't combine these two nodes resulting in having two nodes in the stack 3 . Second column in Table 1 gives this number for both NonInc and RevInc algorithms. Though our algorithm is not fully connected, connectedness of our algorithm is significantly lower than the NonInc algorithm as our algorithm is more incremental. We define waiting time as the number of nodes that need to be shifted to the stack before a dependency between any two nodes in the stack is resolved. In our example sentence, there is a dependency between 'John' and 'likes'. For NonInc, this dependency is resolved only after all the four remaining words in the sentence are shifted. In other words, it has to wait for four more words before this dependency is resolved and hence the waiting time is four. Whereas, in our RevInc algorithm, this dependency is resolved immediately, without waiting for more words to be shifted, and hence the waiting time is zero. The third column in Table 1 gives the waiting time for both the algorithms. Since we compromised incrementality in cases like coordination, waiting time for our RevInc algorithm is not zero but it is significantly lower than the NonInc algorithm and hence more incremental. This property is likely to be crucial for future applications in ASR and SMT language modeling.",
"cite_spans": [
{
"start": 483,
"end": 503,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [
{
"start": 866,
"end": 873,
"text": "Table 1",
"ref_id": "TABREF2"
},
{
"start": 1728,
"end": 1736,
"text": "Table 1",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Connectedness and Waiting Time",
"sec_num": "4.2"
},
{
"text": "We trained the perceptron for both NonInc and RevInc algorithms using the CCGbank training data for 30 iterations, and the models which gave best results on development data are directly used for test data. Table 2 gives the unlabeled precision (UP), recall (UR), F-score (UF) and labeled precision (LP), recall (LR), F-score (LF) results of both NonInc and RevInc approaches on the development data. Last column in the table gives the category accuracy. We used the modified CCGbank for all experiments, including NonInc, for consistent comparisons. For NonInc, the modification decreased unlabeled F-score by 0.45%, without a major difference in labeled F-score. Our incremental algorithm gives 1.39% and 0.47% improvements over the NonInc algorithm in unlabeled and labeled F-scores respectively. For both unlabeled and labeled scores, precision of RevInc is slightly lower than NonInc but the recall of RevInc is much higher than NonInc resulting in a better F-score for RevInc. As NonInc is not incremental and as it uses more context to the right while making a decision, it makes more precise actions. But, on the other hand, if a node is reduced, it is not available for future actions. This is not a problem for our RevInc algorithm which is the reason for higher recall. For example, in the example sentence, 'John likes mangoes from India madly', if the object 'mangoes' is reduced after it got shifted to the stack, then in case of NonInc, the preposition phrase 'from India' can never be attached to 'mangoes'. But, RevInc makes the correct attachment using RRev action. Category accuracy of NonInc is better than RevInc, since NonInc can use more context before taking a complex action and is less prone to error propagation compared to RevInc.",
"cite_spans": [],
"ref_spans": [
{
"start": 207,
"end": 214,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and Analysis",
"sec_num": "4.3"
},
{
"text": "To compare these results in the perspective of Z&C's parser we also trained our NonInc parser with a beam size of 16 similar to Z&C. The second last row in Table 2 (NonInc (beam=16)) shows these results and the last row presents the results from their paper. Results with our implementation of Z&C are 0.65% lower than the published results, possibly due to the modification made in the head rule, and other minor differences like the supertagger beta value. Unlabeled and labeled F-scores of our RevInc parser are lower than these numbers. But, given that our RevInc parser doesn't use any beam, these margins are reasonable.",
"cite_spans": [],
"ref_spans": [
{
"start": 156,
"end": 163,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and Analysis",
"sec_num": "4.3"
},
{
"text": "We also analyzed the label-wise scores of both NonInc and RevInc. In general, NonInc is better in precision and RevInc is better in recall. In the case of verbal arguments ((S\\NP)/NP) and verbal modifiers ((S\\NP)\\(S\\NP)), the F-score of RevInc is better than that of NonInc. But NonInc performed better than RevInc in the case of preposition phrase (PP) attachments ((NP\\NP)/NP, ((S\\NP)\\(S\\NP))/NP). More context is required for better PP attachment which is provided by the fact that NonInc has a context of several unreduced types for the model to work with, whereas RevInc has fewer. Whereas actions like LRev are required to correctly attach the verbal modifiers ('madly') if the subject argument ('John') of the verb ('likes') is reduced early. Table 3 gives the results of these CCG lexical categories. Table 4 : Performance on the test data. *: These results are from their paper.",
"cite_spans": [],
"ref_spans": [
{
"start": 750,
"end": 757,
"text": "Table 3",
"ref_id": "TABREF5"
},
{
"start": 809,
"end": 816,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and Analysis",
"sec_num": "4.3"
},
{
"text": "We also analyzed the performance of the greedy (beam=1) NonInc and RevInc parsers in terms of parsing speed (excluding pos tagger and supertagger time). NonInc and RevInc parse 110 and 125 sentences/second respectively. Despite the complexity of the revealing actions, RevInc is faster than the NonInc. Significant amount of parsing time is spent on the feature extraction step. Features from top four nodes in the stack and their children are extracted for both the algorithms. Since the average connectedness of RevInc and NonInc are 4.62 and 2.15 respectively, on average, all four nodes in the stack are processed for NonInc and only two nodes are processed for RevInc. Because of this there is significant reduction in the feature extraction step for RevInc compared to NonInc. Also, the complex LRev and RRev actions only constituted 5% of the total actions in the parsing process. Table 4 presents the results of our approaches on test data. Our incremental algorithm, RevInc, gives 2.0% and 0.88% improvements over NonInc in unlabeled and labeled F-scores respectively on the test data. Results of RevInc without a beam are reasonably close to the results of Z&C which uses a beam of 16. We compare our results with Incre-mental+Lookahead model of Hassan et al. (2009) . They reported 86.31% unlabeled F-score on test data which is 2.69% lower. Note that these Fscores are not directly comparable since Hassan et al. (2009) use simplified lexicalized CCG categories. Our evaluation is based on CCG dependencies which are different from dependencies in the dependency grammar. Hence, we can't directly compare our results with dependency parsers like Zhang and Nivre (2011) and Honnibal et al. (2013) .",
"cite_spans": [
{
"start": 1256,
"end": 1276,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
},
{
"start": 1411,
"end": 1431,
"text": "Hassan et al. (2009)",
"ref_id": "BIBREF11"
},
{
"start": 1658,
"end": 1680,
"text": "Zhang and Nivre (2011)",
"ref_id": "BIBREF29"
},
{
"start": 1685,
"end": 1707,
"text": "Honnibal et al. (2013)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [
{
"start": 888,
"end": 895,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results and Analysis",
"sec_num": "4.3"
},
{
"text": "We have designed and implemented a new incremental shift-reduce algorithm based on a version of revealing for parsing CCG (Pareschi and Steedman, 1987) . On the standard CCGbank test data, our algorithm achieved improvements of 0.88% and 2.0% in labeled and unlabeled F-scores respectively over the baseline non-incremental shift-reduce algorithm. We achieved this without changing any CCG lexical categories and only changing a single head rule of making the main verb rather than the auxiliary verb the head. Our algorithm models transitions rather than incremental derivations, and hence we don't need an incremental CCGbank. Our approach can therefore be adapted to languages with dependency treebanks, since CCG lexical categories can be easily extracted from dependency treebanks (Cakici, 2005; Ambati et al., 2013) . We also designed new measures of incrementality and showed that our algorithm is more incremental than the standard shift-reduce CCG parsing algorithm.",
"cite_spans": [
{
"start": 122,
"end": 151,
"text": "(Pareschi and Steedman, 1987)",
"ref_id": "BIBREF19"
},
{
"start": 786,
"end": 800,
"text": "(Cakici, 2005;",
"ref_id": "BIBREF2"
},
{
"start": 801,
"end": 821,
"text": "Ambati et al., 2013)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Plan",
"sec_num": "5"
},
{
"text": "We expect to improve our current model in a number of ways. Providing information about lexical category probabilities (Auli and Lopez, 2011) assigned by the supertagger can be useful during parsing. We would like to explore the limited use of a beam to handle lexical ambiguity by only keeping analyses derived from distinct lexical categories in the beam. Following Xu et al. (2014) , we also plan to explore a dynamic oracle strategy. Ultimately, we intend to evaluate the impact of our incremental parser extrinsically in terms of language modeling for SMT or ASR.",
"cite_spans": [
{
"start": 119,
"end": 141,
"text": "(Auli and Lopez, 2011)",
"ref_id": "BIBREF1"
},
{
"start": 368,
"end": 384,
"text": "Xu et al. (2014)",
"ref_id": "BIBREF26"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Plan",
"sec_num": "5"
},
{
"text": "Xu et al. (2014) also obtain CCG dependencies after every action. But they don't have a dependency graph which is updated based on the CCG derivation and used in the CCG parsing (in our case for LRev and RRev actions).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Type-raising followed by forward composition is treated as a single step. Without this, after type-raising, the parser has to check all possible actions before applying forward composition, making it slower.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "This is a case where the dependencies are not true to the CCG grammar, and make our algorithm less incremental than SCH would allow.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "We thank Mike Lewis, Greg Coppola, Francesco Sartorio and Siva Reddy for helpful discussions. We also thank the three anonymous reviewers for their useful suggestions. This work was supported by ERC Advanced Fellowship 249520 GRAMPLUS, EU IST Cognitive Systems IP Xperience and ARC Discovery grant DP 110102506.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Using CCG categories to improve Hindi dependency parsing",
"authors": [],
"year": 2013,
"venue": "Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics",
"volume": "2",
"issue": "",
"pages": "604--609",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bharat Ram Ambati, Tejaswini Deoskar, and Mark Steed- man. 2013. Using CCG categories to improve Hindi dependency parsing. In Proceedings of the 51st An- nual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 604-609, Sofia, Bulgaria.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "A Comparison of Loopy Belief Propagation and Dual Decomposition for Integrated CCG Supertagging and Parsing",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Auli",
"suffix": ""
},
{
"first": "Adam",
"middle": [],
"last": "Lopez",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "470--480",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Auli and Adam Lopez. 2011. A Comparison of Loopy Belief Propagation and Dual Decomposition for Integrated CCG Supertagging and Parsing. In Pro- ceedings of the 49th Annual Meeting of the Associa- tion for Computational Linguistics: Human Language Technologies, pages 470-480, Portland, Oregon, USA, June.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Automatic induction of a CCG grammar for Turkish",
"authors": [
{
"first": "Ruken",
"middle": [],
"last": "Cakici",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the ACL Student Research Workshop",
"volume": "",
"issue": "",
"pages": "73--78",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ruken Cakici. 2005. Automatic induction of a CCG grammar for Turkish. In Proceedings of the ACL Stu- dent Research Workshop, pages 73-78, Ann Arbor, Michigan.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "The importance of supertagging for wide-coverage CCG parsing",
"authors": [
{
"first": "Stephen",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "James",
"middle": [
"R"
],
"last": "Curran",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of COLING-04",
"volume": "",
"issue": "",
"pages": "282--288",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stephen Clark and James R. Curran. 2004. The impor- tance of supertagging for wide-coverage CCG parsing. In Proceedings of COLING-04, pages 282-288.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Widecoverage efficient statistical parsing with CCG and log-linear models",
"authors": [
{
"first": "Stephen",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "James",
"middle": [
"R"
],
"last": "Curran",
"suffix": ""
}
],
"year": 2007,
"venue": "Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stephen Clark and James R. Curran. 2007. Wide- coverage efficient statistical parsing with CCG and log-linear models. Computational Linguistics, 33.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Building Deep Dependency Structures using a Wide-Coverage CCG Parser",
"authors": [
{
"first": "Stephen",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "Julia",
"middle": [],
"last": "Hockenmaier",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2002,
"venue": "ACL",
"volume": "",
"issue": "",
"pages": "327--334",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stephen Clark, Julia Hockenmaier, and Mark Steedman. 2002. Building Deep Dependency Structures using a Wide-Coverage CCG Parser. In ACL, pages 327-334.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Incremental Parsing with the Perceptron Algorithm",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
},
{
"first": "Brian",
"middle": [],
"last": "Roark",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 42nd Meeting of the Association for Computational Linguistics (ACL'04), Main Volume",
"volume": "",
"issue": "",
"pages": "111--118",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins and Brian Roark. 2004. Incremental Parsing with the Perceptron Algorithm. In Proceed- ings of the 42nd Meeting of the Association for Com- putational Linguistics (ACL'04), Main Volume, pages 111-118, Barcelona, Spain, July.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Discriminative training methods for hidden Markov models: theory and experiments with perceptron algorithms",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the conference on Empirical methods in natural language processing, EMNLP '02",
"volume": "",
"issue": "",
"pages": "1--8",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 2002. Discriminative training methods for hidden Markov models: theory and experiments with perceptron algorithms. In Proceedings of the con- ference on Empirical methods in natural language pro- cessing, EMNLP '02, pages 1-8.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Ellipsis and higher-order unification",
"authors": [
{
"first": "Mary",
"middle": [],
"last": "Dalrymple",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Stuart",
"suffix": ""
},
{
"first": "Fernando Cn",
"middle": [],
"last": "Shieber",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Pereira",
"suffix": ""
}
],
"year": 1991,
"venue": "Linguistics and philosophy",
"volume": "14",
"issue": "4",
"pages": "399--452",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mary Dalrymple, Stuart M Shieber, and Fernando CN Pereira. 1991. Ellipsis and higher-order unification. Linguistics and philosophy, 14(4):399-452.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Efficient Normal-Form Parsing for Combinatory Categorial Grammar",
"authors": [
{
"first": "Jason",
"middle": [],
"last": "Eisner",
"suffix": ""
}
],
"year": 1996,
"venue": "Proceedings of the 34th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "79--86",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jason Eisner. 1996. Efficient Normal-Form Parsing for Combinatory Categorial Grammar. In Proceedings of the 34th Annual Meeting of the Association for Com- putational Linguistics, pages 79-86, Santa Cruz, Cali- fornia, USA, June.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "A Dynamic Oracle for Arc-Eager Dependency Parsing",
"authors": [
{
"first": "Yoav",
"middle": [],
"last": "Goldberg",
"suffix": ""
},
{
"first": "Joakim",
"middle": [],
"last": "Nivre",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of COLING 2012",
"volume": "",
"issue": "",
"pages": "959--976",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yoav Goldberg and Joakim Nivre. 2012. A Dynamic Oracle for Arc-Eager Dependency Parsing. In Pro- ceedings of COLING 2012, pages 959-976, Mumbai, India, December.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Lexicalized Semi-Incremental Dependency Parsing",
"authors": [
{
"first": "Hany",
"middle": [],
"last": "Hassan",
"suffix": ""
},
{
"first": "Khalil",
"middle": [],
"last": "Sima'an",
"suffix": ""
},
{
"first": "Andy",
"middle": [],
"last": "Way",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the Recent Advances in NLP (RANLP'09)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hany Hassan, Khalil Sima'an, and Andy Way. 2009. Lexicalized Semi-Incremental Dependency Parsing. In Proceedings of the Recent Advances in NLP (RANLP'09), Borovets, Bulgaria.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Generative models for statistical parsing with Combinatory Categorial Grammar",
"authors": [
{
"first": "Julia",
"middle": [],
"last": "Hockenmaier",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 40th Annual Meeting on Association for Computational Linguistics, ACL '02",
"volume": "",
"issue": "",
"pages": "335--342",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Julia Hockenmaier and Mark Steedman. 2002. Gener- ative models for statistical parsing with Combinatory Categorial Grammar. In Proceedings of the 40th An- nual Meeting on Association for Computational Lin- guistics, ACL '02, pages 335-342, Philadelphia, Penn- sylvania.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "CCGbank: A Corpus of CCG Derivations and Dependency Structures Extracted from the Penn Treebank",
"authors": [
{
"first": "Julia",
"middle": [],
"last": "Hockenmaier",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2007,
"venue": "Computational Linguistics",
"volume": "33",
"issue": "3",
"pages": "355--396",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Julia Hockenmaier and Mark Steedman. 2007. CCG- bank: A Corpus of CCG Derivations and Dependency Structures Extracted from the Penn Treebank. Com- putational Linguistics, 33(3):355-396.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "A Non-Monotonic Arc-Eager Transition System for Dependency Parsing",
"authors": [
{
"first": "Matthew",
"middle": [],
"last": "Honnibal",
"suffix": ""
},
{
"first": "Yoav",
"middle": [],
"last": "Goldberg",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Johnson",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the Seventeenth Conference on Computational Natural Language Learning",
"volume": "",
"issue": "",
"pages": "163--172",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Matthew Honnibal, Yoav Goldberg, and Mark John- son. 2013. A Non-Monotonic Arc-Eager Transition System for Dependency Parsing. In Proceedings of the Seventeenth Conference on Computational Natural Language Learning, pages 163-172, Sofia, Bulgaria, August.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Inducing probabilistic CCG grammars from logical form with higher-order unification",
"authors": [
{
"first": "Tom",
"middle": [],
"last": "Kwiatkowski",
"suffix": ""
},
{
"first": "Luke",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
},
{
"first": "Sharon",
"middle": [],
"last": "Goldwater",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2010,
"venue": "Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "1223--1233",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tom Kwiatkowski, Luke Zettlemoyer, Sharon Goldwa- ter, and Mark Steedman. 2010. Inducing probabilistic CCG grammars from logical form with higher-order unification. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1223-1233, Cambridge, MA, October.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "A* CCG Parsing with a Supertag-factored Model",
"authors": [
{
"first": "Mike",
"middle": [],
"last": "Lewis",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mike Lewis and Mark Steedman. 2014. A* CCG Pars- ing with a Supertag-factored Model. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, Doha, Qatar, October.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Linguistic structure and speech shadowing at very short latencies",
"authors": [
{
"first": "W",
"middle": [],
"last": "Marslen-Wilson",
"suffix": ""
}
],
"year": 1973,
"venue": "Nature",
"volume": "244",
"issue": "",
"pages": "522--533",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "W. Marslen-Wilson. 1973. Linguistic structure and speech shadowing at very short latencies. Nature., 244:522-533.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Maltparser: A languageindependent system for data-driven dependency parsing",
"authors": [
{
"first": "Joakim",
"middle": [],
"last": "Nivre",
"suffix": ""
},
{
"first": "Johan",
"middle": [],
"last": "Hall",
"suffix": ""
},
{
"first": "Jens",
"middle": [],
"last": "Nilsson",
"suffix": ""
},
{
"first": "Atanas",
"middle": [],
"last": "Chanev",
"suffix": ""
},
{
"first": "G\u00fclsen",
"middle": [],
"last": "Eryigit",
"suffix": ""
},
{
"first": "Sandra",
"middle": [],
"last": "K\u00fcbler",
"suffix": ""
},
{
"first": "Svetoslav",
"middle": [],
"last": "Marinov",
"suffix": ""
},
{
"first": "Erwin",
"middle": [],
"last": "Marsi",
"suffix": ""
}
],
"year": 2007,
"venue": "Natural Language Engineering",
"volume": "13",
"issue": "2",
"pages": "95--135",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Joakim Nivre, Johan Hall, Jens Nilsson, Atanas Chanev, G\u00fclsen Eryigit, Sandra K\u00fcbler, Svetoslav Marinov, and Erwin Marsi. 2007. Maltparser: A language- independent system for data-driven dependency pars- ing. Natural Language Engineering, 13(2):95-135.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "A lazy way to chart-parse with categorial grammars",
"authors": [
{
"first": "Remo",
"middle": [],
"last": "Pareschi",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 1987,
"venue": "Proceedings of the 25th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "81--88",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Remo Pareschi and Mark Steedman. 1987. A lazy way to chart-parse with categorial grammars. In Proceed- ings of the 25th Annual Meeting of the Association for Computational Linguistics, pages 81-88, Stanford, California, USA, July.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Probabilistic top-down parsing and language modeling",
"authors": [
{
"first": "Brian",
"middle": [],
"last": "Roark",
"suffix": ""
}
],
"year": 2001,
"venue": "Computational Linguistics",
"volume": "27",
"issue": "",
"pages": "249--276",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Brian Roark. 2001. Probabilistic top-down parsing and language modeling. Computational Linguistics, 27:249-276.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Incremental Tree Substitution Grammar for Parsing and Sentence Prediction",
"authors": [
{
"first": "Federico",
"middle": [],
"last": "Sangati",
"suffix": ""
},
{
"first": "Frank",
"middle": [],
"last": "Keller",
"suffix": ""
}
],
"year": 2013,
"venue": "Transactions of the Association for Computational Linguistics (TACL)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Federico Sangati and Frank Keller. 2013. Incremen- tal Tree Substitution Grammar for Parsing and Sen- tence Prediction. In Transactions of the Association for Computational Linguistics (TACL).",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "A Transition-Based Dependency Parser Using a Dynamic Parsing Strategy",
"authors": [
{
"first": "Francesco",
"middle": [],
"last": "Sartorio",
"suffix": ""
},
{
"first": "Giorgio",
"middle": [],
"last": "Satta",
"suffix": ""
},
{
"first": "Joakim",
"middle": [],
"last": "Nivre",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the 51st",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Francesco Sartorio, Giorgio Satta, and Joakim Nivre. 2013. A Transition-Based Dependency Parser Using a Dynamic Parsing Strategy. In Proceedings of the 51st",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Annual Meeting of the Association for Computational Linguistics",
"authors": [],
"year": null,
"venue": "",
"volume": "1",
"issue": "",
"pages": "135--144",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 135-144, Sofia, Bulgaria, August.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "The Syntactic Process",
"authors": [
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2000,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mark Steedman. 2000. The Syntactic Process. MIT Press, Cambridge, MA, USA.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Language modeling using a statistical dependency grammar parser",
"authors": [
{
"first": "Wen",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Mary",
"middle": [],
"last": "Harper",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the International Workshop on Automatic Speech Recognition and Understanding, US Virgin Islands",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wen Wang and Mary Harper. 2003. Language modeling using a statistical dependency grammar parser. In Pro- ceedings of the International Workshop on Automatic Speech Recognition and Understanding, US Virgin Is- lands.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Shift-Reduce CCG Parsing with a Dependency Model",
"authors": [
{
"first": "Wenduan",
"middle": [],
"last": "Xu",
"suffix": ""
},
{
"first": "Stephen",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "Yue",
"middle": [],
"last": "Zhang",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics",
"volume": "1",
"issue": "",
"pages": "218--227",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wenduan Xu, Stephen Clark, and Yue Zhang. 2014. Shift-Reduce CCG Parsing with a Dependency Model. In Proceedings of the 52nd Annual Meeting of the As- sociation for Computational Linguistics (Volume 1: Long Papers), pages 218-227, Baltimore, Maryland, June.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Statistical Dependency Analysis with Support Vector Machines",
"authors": [
{
"first": "Hiroyasu",
"middle": [],
"last": "Yamada",
"suffix": ""
},
{
"first": "Yuji",
"middle": [],
"last": "Matsumoto",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of IWPT",
"volume": "",
"issue": "",
"pages": "195--206",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hiroyasu Yamada and Yuji Matsumoto. 2003. Statistical Dependency Analysis with Support Vector Machines. In In Proceedings of IWPT, pages 195-206.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Shift-Reduce CCG Parsing",
"authors": [
{
"first": "Yue",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Stephen",
"middle": [],
"last": "Clark",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "683--692",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yue Zhang and Stephen Clark. 2011. Shift-Reduce CCG Parsing. In Proceedings of the 49th Annual Meet- ing of the Association for Computational Linguistics: Human Language Technologies, pages 683-692, Port- land, Oregon, USA, June.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Transition-based Dependency Parsing with Rich Non-local Features",
"authors": [
{
"first": "Yue",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Joakim",
"middle": [],
"last": "Nivre",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "188--193",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yue Zhang and Joakim Nivre. 2011. Transition-based Dependency Parsing with Rich Non-local Features. In Proceedings of the 49th Annual Meeting of the Asso- ciation for Computational Linguistics: Human Lan- guage Technologies, pages 188-193, Portland, Ore- gon, USA.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"num": null,
"uris": null,
"text": "Normal form CCG derivation.",
"type_str": "figure"
},
"FIGREF1": {
"num": null,
"uris": null,
"text": "RevInc -Sequence of actions with parser configuration and the corresponding dependency graph.",
"type_str": "figure"
},
"FIGREF2": {
"num": null,
"uris": null,
"text": "RRev and LRev actions.",
"type_str": "figure"
},
"TABREF2": {
"text": "Connectedness and waiting time.",
"content": "<table/>",
"type_str": "table",
"html": null,
"num": null
},
"TABREF3": {
"text": "Performance on the development data. *: These results are from the Z&C paper.",
"content": "<table><tr><td/><td>UP</td><td>UR</td><td>UF</td><td>LP</td><td>LR</td><td>LF</td><td>Cat Acc.</td></tr><tr><td>NonInc (beam=1)</td><td colspan=\"6\">92.57 82.60 87.30 85.12 75.96 80.28</td><td>91.10</td></tr><tr><td>RevInc (beam=1)</td><td colspan=\"6\">91.62 85.94 88.69 83.42 78.25 80.75</td><td>90.87</td></tr><tr><td colspan=\"7\">NonInc (beam=16) 92.71 89.66 91.16 85.78 82.96 84.35</td><td>92.51</td></tr><tr><td>Z&C (beam=16)*</td><td>-</td><td>-</td><td>-</td><td colspan=\"3\">87.15 82.95 85.00</td><td>92.77</td></tr><tr><td>Table 2:</td><td/><td/><td/><td/><td/><td/></tr></table>",
"type_str": "table",
"html": null,
"num": null
},
"TABREF5": {
"text": "Label-wise F-score of RevInc and NonInc parsers (both with beam=1). Argument slots in the relation are in bold. 92.45 82.16 87.00 85.59 76.06 80.55 91.39 RevInc (beam=1) 91.83 86.35 89.00 84.02 79.00 81.43 91.17 NonInc (beam=16) 92.68 89.57 91.10 86.20 83.32 84.74",
"content": "<table><tr><td/><td>UP</td><td>UR</td><td>UF</td><td>LP</td><td>LR</td><td>LF</td><td>Cat Acc.</td></tr><tr><td>NonInc (beam=1)</td><td/><td/><td/><td/><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>92.70</td></tr><tr><td>Z&C (beam=16)*</td><td>-</td><td>-</td><td>-</td><td colspan=\"3\">87.43 83.61 85.48</td><td>93.12</td></tr><tr><td>Hassan et al. 09*</td><td>-</td><td>-</td><td>86.31</td><td>-</td><td>-</td><td>-</td><td>-</td></tr></table>",
"type_str": "table",
"html": null,
"num": null
}
}
}
} |