File size: 99,780 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 | {
"paper_id": "P13-1022",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T09:34:12.451122Z"
},
"title": "Adapting Discriminative Reranking to Grounded Language Learning",
"authors": [
{
"first": "Joohyun",
"middle": [],
"last": "Kim",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Texas at Austin Austin",
"location": {
"postCode": "78701",
"region": "TX",
"country": "USA"
}
},
"email": ""
},
{
"first": "Raymond",
"middle": [
"J"
],
"last": "Mooney",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Texas at Austin Austin",
"location": {
"postCode": "78701",
"region": "TX",
"country": "USA"
}
},
"email": "mooney@cs.utexas.edu"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "We adapt discriminative reranking to improve the performance of grounded language acquisition, specifically the task of learning to follow navigation instructions from observation. Unlike conventional reranking used in syntactic and semantic parsing, gold-standard reference trees are not naturally available in a grounded setting. Instead, we show how the weak supervision of response feedback (e.g. successful task completion) can be used as an alternative, experimentally demonstrating that its performance is comparable to training on gold-standard parse trees.",
"pdf_parse": {
"paper_id": "P13-1022",
"_pdf_hash": "",
"abstract": [
{
"text": "We adapt discriminative reranking to improve the performance of grounded language acquisition, specifically the task of learning to follow navigation instructions from observation. Unlike conventional reranking used in syntactic and semantic parsing, gold-standard reference trees are not naturally available in a grounded setting. Instead, we show how the weak supervision of response feedback (e.g. successful task completion) can be used as an alternative, experimentally demonstrating that its performance is comparable to training on gold-standard parse trees.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Grounded language acquisition involves learning to comprehend and/or generate language by simply observing its use in a naturally occurring context in which the meaning of a sentence is grounded in perception and/or action (Roy, 2002; Yu and Ballard, 2004; Gold and Scassellati, 2007; Chen et al., 2010) . B\u00f6rschinger et al. (2011) introduced an approach that reduces grounded language learning to unsupervised probabilistic context-free grammar (PCFG) induction and demonstrated its effectiveness on the task of sportscasting simulated robot soccer games. Subsequently, Kim and Mooney (2012) extended their approach to make it tractable for the more complex problem of learning to follow natural-language navigation instructions from observations of humans following such instructions in a virtual environment (Chen and Mooney, 2011) . The observed sequence of actions provides very weak, ambiguous supervision for learning instructional language since there are many possible ways to describe the same execution path. Although their approach improved accuracy on the navigation task compared to the original work of Chen and Mooney (2011) , it was still far from human performance.",
"cite_spans": [
{
"start": 223,
"end": 234,
"text": "(Roy, 2002;",
"ref_id": "BIBREF21"
},
{
"start": 235,
"end": 256,
"text": "Yu and Ballard, 2004;",
"ref_id": "BIBREF25"
},
{
"start": 257,
"end": 284,
"text": "Gold and Scassellati, 2007;",
"ref_id": "BIBREF12"
},
{
"start": 285,
"end": 303,
"text": "Chen et al., 2010)",
"ref_id": "BIBREF3"
},
{
"start": 306,
"end": 331,
"text": "B\u00f6rschinger et al. (2011)",
"ref_id": "BIBREF0"
},
{
"start": 571,
"end": 592,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
},
{
"start": 811,
"end": 834,
"text": "(Chen and Mooney, 2011)",
"ref_id": "BIBREF2"
},
{
"start": 1118,
"end": 1140,
"text": "Chen and Mooney (2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Since their system employs a generative model, discriminative reranking (Collins, 2000) could potentially improve its performance. By training a discriminative classifier that uses global features of complete parses to identify correct interpretations, a reranker can significantly improve the accuracy of a generative model. Reranking has been successfully employed to improve syntactic parsing (Collins, 2002b) , semantic parsing (Lu et al., 2008; Ge and Mooney, 2006) , semantic role labeling (Toutanova et al., 2005) , and named entity recognition (Collins, 2002c) . Standard reranking requires gold-standard interpretations (e.g. parse trees) to train the discriminative classifier. However, grounded language learning does not provide gold-standard interpretations for the training examples. Only the ambiguous perceptual context of the utterance is provided as supervision. For the navigation task, this supervision consists of the observed sequence of actions taken by a human when following an instruction. Therefore, it is impossible to directly apply conventional discriminative reranking to such problems. We show how to adapt reranking to work with such weak supervision. Instead of using gold-standard annotations to determine the correct interpretations, we simply prefer interpretations of navigation instructions that, when executed in the world, actually reach the intended destination. Additionally, we extensively revise the features typically used in parse reranking to work with the PCFG approach to grounded language learning.",
"cite_spans": [
{
"start": 72,
"end": 87,
"text": "(Collins, 2000)",
"ref_id": "BIBREF6"
},
{
"start": 396,
"end": 412,
"text": "(Collins, 2002b)",
"ref_id": null
},
{
"start": 432,
"end": 449,
"text": "(Lu et al., 2008;",
"ref_id": "BIBREF19"
},
{
"start": 450,
"end": 470,
"text": "Ge and Mooney, 2006)",
"ref_id": "BIBREF11"
},
{
"start": 496,
"end": 520,
"text": "(Toutanova et al., 2005)",
"ref_id": "BIBREF23"
},
{
"start": 552,
"end": 568,
"text": "(Collins, 2002c)",
"ref_id": "BIBREF9"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The rest of the paper is organized as follows: Section 2 reviews the navigation task and the PCFG approach to grounded language learning. Section 3 presents our modified approach to reranking and Section 4 describes the novel features used to evaluate parses. Section 5 experimentally evaluates the approach comparing to sev-(a) Sample virtual world of hallways with varying tiles, wallpapers, and landmark objects indicated by letters (e.g. \"H\" for hat-rack) and illustrating a sample path taken by a human follower.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "(b) A sample natural language instruction and its formal landmarks plan for the path illustrated above. The subset corresponding to the correct formal plan is shown in bold. eral baselines. Finally, Section 6 describes related work, Section 7 discusses future work, and Section 8 concludes.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "We address the navigation learning task introduced by Chen and Mooney (2011) . The goal is to interpret natural-language (NL) instructions in a virtual environment, thereby allowing a simulated robot to navigate to a specified location. Figure 1a shows a sample path executed by a human following the instruction in Figure 1b . Given no prior linguistic knowledge, the task is to learn to interpret such instructions by simply observing humans follow sample directions. Formally speaking, given training examples of the form (e i , a i , w i ), where e i is an NL instruction, a i is an executed action sequence for the instruction, and w i is the initial world state, we want to learn to produce an appropriate action sequence a j given a novel (e j , w j ).",
"cite_spans": [
{
"start": 54,
"end": 76,
"text": "Chen and Mooney (2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [
{
"start": 237,
"end": 246,
"text": "Figure 1a",
"ref_id": "FIGREF0"
},
{
"start": 316,
"end": 325,
"text": "Figure 1b",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Navigation Task",
"sec_num": "2.1"
},
{
"text": "More specifically, one must learn a semantic parser that produces a plan p j using a formal meaning representation (MR) language introduced by Chen and Mooney (2011) . This plan is then executed by a simulated robot in a virtual environment. The MARCO system, introduced by MacMahon et al. 2006, executes the formal plan, flexibly adapting to situations encountered during execution and producing the action sequence a j . During learning, Chen and Mooney construct a landmarks plan c i for each training example, which includes the complete context observed in the world-state resulting from each observed action. The correct plan, p i , (which is latent and must be inferred) is assumed to be composed from a subset of the components in the corresponding landmarks plan. The landmarks and correct plans for a sample instruction are shown in Figure 1b .",
"cite_spans": [
{
"start": 143,
"end": 165,
"text": "Chen and Mooney (2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [
{
"start": 843,
"end": 852,
"text": "Figure 1b",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Navigation Task",
"sec_num": "2.1"
},
{
"text": "The baseline generative model we use for reranking employs the unsupervised PCFG induction approach introduced by Kim and Mooney (2012) . This model is, in turn, based on the earlier model of B\u00f6rschinger et al. (2011) , which transforms the grounded language learning into unsupervised PCFG induction. The general approach uses grammar-formulation rules which construct CFG productions that form a grammar that effectively maps NL sentences to formal meaning representations (MRs) encoded in its nonterminals. After using Expectation-Maximization (EM) to estimate the parameters for these productions using the ambiguous supervision provided by the groundedlearning setting, it produces a PCFG whose most probable parse for a sentence encodes its correct semantic interpretation. Unfortunately, the initial approach of B\u00f6rschinger et al. (2011) produces explosively large grammars when applied to more complex problems, such as our navigation task. Therefore, Kim and Mooney enhanced their approach to use a previously learned semantic lexicon to reduce the induced grammar to a tractable size. They also altered the processes for constructing productions and mapping parse trees to MRs in order to make the construction of semantic interpretations more compositional and allow the efficient construction of more complex representa- ",
"cite_spans": [
{
"start": 114,
"end": 135,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
},
{
"start": 192,
"end": 217,
"text": "B\u00f6rschinger et al. (2011)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "PCFG Induction for Grounded Language Learning",
"sec_num": "2.2"
},
{
"text": "The resulting PCFG can be used to produce a set of most-probable interpretations of instructional sentences for the navigation task. Our proposed reranking model is used to discriminatively reorder the top parses produced by this generative model. A simplified version of a sample parse tree for Kim and Mooney's model is shown in Figure 2 .",
"cite_spans": [],
"ref_spans": [
{
"start": 331,
"end": 339,
"text": "Figure 2",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "tions.",
"sec_num": null
},
{
"text": "In reranking, a baseline generative model is first trained and generates a set of candidate outputs for each training example. Next, a second conditional model is trained which uses global features to rescore the candidates. Reranking using an averaged perceptron (Collins, 2002a) has been successfully applied to a variety of NLP tasks. Therefore, we modify it to rerank the parse trees generated by Kim and Mooney (2012) 's model. The approach requires three subcomponents: 1) a GEN function that returns the list of top n candidate parse trees for each NL sentence produced by the generative model, 2) a feature function \u03a6 that maps a NL sentence, e, and a parse tree, y, into a real-valued feature vector \u03a6(e, y) \u2208 R d , and 3) a reference parse tree that is compared to the highest-scoring parse tree during training.",
"cite_spans": [
{
"start": 264,
"end": 280,
"text": "(Collins, 2002a)",
"ref_id": "BIBREF7"
},
{
"start": 401,
"end": 422,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "However, grounded language learning tasks, such as our navigation task, do not provide reference parse trees for training examples. Instead, our modified model replaces the gold-standard reference parse with the \"pseudo-gold\" parse tree ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "InitializeW = 0 3: for t = 1...T, i = 1...n do 4: y i = arg max y\u2208GEN(e i ) \u03a6(e i , y) \u2022W 5: if y i = y * i then 6:W =W + \u03a6(e i , y * i ) \u2212 \u03a6(e i , y i ) 7: end if 8:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "end for 9: end procedure whose derived MR plan is most successful at getting to the desired goal location. Thus, the third component in our reranking model becomes an evaluation function EXEC that maps a parse tree y into a real number representing the success rate (w.r.t. successfully reaching the intended destination) of the derived MR plan m composed from y.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "Additionally, we improve the perceptron training algorithm by using multiple reference parses to update the weight vectorW .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "Although we determine the pseudo-gold reference tree to be the candidate parse y * such that y * = arg max y\u2208GEN(e) EXEC(y), it may not actually be the correct parse for the sentence. Other parses may contain useful information for learning, and therefore we devise a way to update weights using all candidate parses whose successful execution rate is greater than the parse preferred by the currently learned model.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Modified Reranking Algorithm",
"sec_num": "3"
},
{
"text": "To circumvent the need for gold-standard reference parses, we select a pseudo-gold parse from the candidates produced by the GEN function. In a similar vein, when reranking semantic parses, Ge and Mooney (2006) chose as a reference parse the one which was most similar to the gold-standard semantic annotation. However, in the navigation task, the ultimate goal is to generate a plan that, when actually executed in the virtual environment, leads to the desired destination. Therefore, the pseudo-gold reference is chosen as the candidate parse that produces the MR plan with the great-est execution success. This requires an external module that evaluates the execution accuracy of the candidate parses. For the navigation task, we use the MARCO (MacMahon et al., 2006) execution module, which is also used to evaluate how well the overall system learns to follow directions (Chen and Mooney, 2011) . Since MARCO is nondeterministic when executing underspecified plans, we execute each candidate plan 10 times, and its execution rate is the percentage of trials in which it reaches the correct destination. When there are multiple candidate parses tied for the highest execution rate, the one assigned the largest probability by the baseline model is selected. Our modified averaged perceptron procedure with such a response-based update is shown in Algorithm 1.",
"cite_spans": [
{
"start": 190,
"end": 210,
"text": "Ge and Mooney (2006)",
"ref_id": "BIBREF11"
},
{
"start": 741,
"end": 770,
"text": "MARCO (MacMahon et al., 2006)",
"ref_id": null
},
{
"start": 876,
"end": 899,
"text": "(Chen and Mooney, 2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Response-Based Weight Updates",
"sec_num": "3.1"
},
{
"text": "One additional issue must be addressed when computing the output of the GEN function. The final plan MRs are produced from parse trees using compositional semantics (see Kim and Mooney (2012) for details). Consequently, the n-best parse trees for the baseline model do not necessarily produce the n-best distinct plans, since many parses can produce the same plan. Therefore, we adapt the GEN function to produce the n best distinct plans rather than the n best parses. This may require examining many more than the n best parses, because many parses have insignificant differences that do not affect the final plan. The score assigned to a plan is the probability of the most probable parse that generates that plan. In order to efficiently compute the n best plans, we modify the exact n-best parsing algorithm developed by Huang and Chiang (2005) . The modified algorithm ensures that each plan in the computed n best list produces a new distinct plan.",
"cite_spans": [
{
"start": 170,
"end": 191,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
},
{
"start": 826,
"end": 849,
"text": "Huang and Chiang (2005)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Response-Based Weight Updates",
"sec_num": "3.1"
},
{
"text": "Typically, when used for reranking, the averaged perceptron updates its weights using the featurevector difference between the current best predicted candidate and the gold-standard reference (line 6 in Algorithm 1). In our initial modified version, we replaced the gold-standard reference parse with the pseudo-gold reference, which has the highest execution rate amongst all candidate parses. However, this ignores all other candidate parses during perceptron training. However, it is not ideal to regard other candidate parses as \"useless.\" There may be multiple candidate parses with the same maximum execution rate, and even can-didates with lower execution rates could represent the correct plan for the instruction given the weak, indirect supervision provided by the observed sequence of human actions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "Therefore, we also consider a further modification of the averaged perceptron algorithm which updates its weights using multiple candidate parses. Instead of only updating the weights with the single difference between the predicted and pseudo-gold parses, the weight vectorW is updated with the sum of feature-vector differences between the current predicted candidate and all other candidates that have a higher execution rate. Formally, in this version, we replace lines 5-6 of Algorithm 1 with:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "1: for all y \u2208 GEN(e i ) where y = y i and EXEC(y)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "> EXEC(y i ) do 2:W =W + (EXEC(y) \u2212 EXEC(y i )) \u00d7(\u03a6(e i , y) \u2212 \u03a6(e i , y i )) 3: end for",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "where EXEC(y) is the execution rate of the MR plan m derived from parse tree y.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "In the experiments below, we demonstrate that, by exploiting multiple reference parses, this new update rule increases the execution accuracy of the final system. Intuitively, this approach gathers additional information from all candidate parses with higher execution accuracy when learning the discriminative reranker. In addition, as shown in line 2 of the algorithm above, it uses the difference in execution rates between a candidate and the currently preferred parse to weight the update to the parameters for that candidate. This allows more effective plans to have a larger impact on the learned model in each iteration.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Weight Updates Using Multiple Parses",
"sec_num": "3.2"
},
{
"text": "This section describes the features \u03a6 extracted from parses produced by the generative model and used to rerank the candidates.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Reranking Features",
"sec_num": "4"
},
{
"text": "The base features adapt those used in previous reranking methods, specifically those of Collins (2002a) , Lu et al. (2008) , and Ge and Mooney (2006) , which are directly extracted from parse trees. In addition, we also include the log probability of the parse tree as an additional feature. Figure 3 shows a sample full parse tree from our baseline model, which is used when explaining the L1: Turn(LEFT), Verify(front : SOFA, back : EASEL), Travel(steps : 2), Verify(at : SOFA), Turn(RIGHT) Figure 3 : Sample full parse tree for the sentence \"Turn left and find the soft then turn around the corner\" used to explain reranking features. Nonterminals representing MR plan components are shown, which are labeled L 1 to L 6 for ease of reference. Additional nonterminals such as P hrase, P h, P hX, and W ord are subsidiary ones for generating NL words from MR nonterminals. They are also shown in order to represent the entire process of how parse trees are constructed (for details, refer to Kim and Mooney (2012) ).",
"cite_spans": [
{
"start": 88,
"end": 103,
"text": "Collins (2002a)",
"ref_id": "BIBREF7"
},
{
"start": 106,
"end": 122,
"text": "Lu et al. (2008)",
"ref_id": "BIBREF19"
},
{
"start": 129,
"end": 149,
"text": "Ge and Mooney (2006)",
"ref_id": "BIBREF11"
},
{
"start": 993,
"end": 1014,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [
{
"start": 292,
"end": 300,
"text": "Figure 3",
"ref_id": null
},
{
"start": 493,
"end": 501,
"text": "Figure 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "L6: Turn() P hrase L 6 W ord L 6 corner P hX L 6 W ord \u2205 the P hX L 6 W ord L 6 around P hX L 6 W ord L 6 turn P hX L 6 W ord \u2205 then L3: Travel(steps : 2), Verify(at : SOFA), Turn(RIGHT) L5: Travel(), Verify(at : SOFA) P hrase L 5 W ord L 5 sofa P hX L 5 W ord \u2205 the P hX L 5 W ord L 5 find L2: Turn(LEFT), Verify(front : SOFA) L4: Turn(LEFT) P hrase L 4 W ord \u2205 and P h L 4 W ord L 4 left P hX L 4 W ord L 4 Turn",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "reranking features below, each illustrated by an example. a) PCFG Rule. Indicates whether a particular PCFG rule is used in the parse tree:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 1 \u21d2 L 2 L 3 ) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "b) Grandparent PCFG Rule. Indicates whether a particular PCFG rule as well as the nonterminal above it is used in the parse tree:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 3 \u21d2 L 5 L 6 |L1) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "c) Long-range Unigram. Indicates whether a nonterminal has a given NL word below it in the parse tree: f (L 2 ; left) = 1 and f (L 4 ; turn) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "d) Two-level Long-range Unigram. Indicates whether a nonterminal has a child nonterminal which eventually generates a NL word in the parse tree: f (L 4 ; left|L 2 ) = 1 e) Unigram. Indicates whether a nonterminal produces a given child nonterminal or terminal NL word in the parse tree:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 1 \u2192 L 2 ) = 1 and f (L 1 \u2192 L 3 ) = 1. f) Grandparent Unigram.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "Indicates whether a nonterminal has a given child nonterminal/terminal below it, as well as a given parent nonterminal:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 2 \u2192 L 4 |L 1 ) = 1 g) Bigram.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "Indicates whether a given bigram of nonterminal/terminals occurs for given a parent nonterminal:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 1 \u2192 L 2 : L 3 ) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "h) Grandparent Bigram. Same as Bigram, but also includes the nonterminal above the parent nonterminal:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "f (L 3 \u2192 L 5 : L 6 |L 1 ) = 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "i) Log-probability of Parse Tree. Certainty assigned by the base generative model.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Base Features",
"sec_num": "4.1"
},
{
"text": "The base features above generally include nonterminal symbols used in the parse tree. In the grounded PCFG model, nonterminals are named after components of the semantic representations (MRs), which are complex and numerous. There are 2,500 nonterminals in the grammar constructed for the navigation data, most of which are very specific and rare. This results in a very large, sparse feature space which can easily lead the reranking model to over-fit the training data and prevent it from generalizing properly.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicate-Only Features",
"sec_num": "4.2"
},
{
"text": "Therefore, we also tried constructing more general features that are less sparse. First, we construct generalized versions of the base features in which nonterminal symbols use only predicate names and omit their arguments. In the navigation task, action arguments frequently contain redundant, rarely used information. In particular, the interleaving verification steps frequently include many details that are never actually mentioned in the NL instructions. For instance, a nonterminal for the MR Turn(LEFT), Verify(at:SOFA,front:EASEL), Travel(steps:3) is transformed into the predicate-only form Turn(), Verify(), Travel() , and then used to construct more general versions of the base features described in the previous section. Second, another version of the base features are constructed in which nonterminal symbols include action arguments but omit all interleaving verification steps. This is a somewhat more conservative simplification of the nonterminal symbols. Although verification steps sometimes help interpret the actions and their surrounding context, they frequently cause the nonterminal symbols to become unnecessarily complex and specific.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicate-Only Features",
"sec_num": "4.2"
},
{
"text": "Finally, another feature group which we utilize captures whether a particular atomic action in a nonterminal \"descends\" into one of its child nonterminals or not. An atomic action consists of a predicate and its arguments, e.g. Turn(LEFT), Travel(steps:2), or Verify(at:SOFA). When an atomic action descends into lower nonterminals in a parse tree, it indicates that it is mentioned in the NL instruction and is therefore important. Below are several feature types related to descended actions that are used in our reranking model: a) Descended Action. Indicates whether a given atomic action in a nonterminal descends to the next level. In Figure 3 The navigation data was collected by MacMahon et al. (2006) , and includes English instructions and human follower data. 1 The data contains 706 route instructions for three virtual worlds. The instructions were produced by six instructors for 126 unique starting and ending location pairs over the three maps. Each instruction is annotated with 1 to 15 human follower traces with an average of 10.4 actions per instruction. Each instruction contains an average of 5.0 sentences each with an average of 7.8 words. Chen and Mooney (2011) constructed a version of the data in which each sentence is annotated with the actions taken by the majority of followers when responding to this sentence. This single-sentence version is used for training. Manually annotated \"gold standard\" formal plans for each sentence are used for evaluation purposes only. We followed the same experimental methodology as Kim and Mooney (2012) and Chen and Mooney (2011) . We performed \"leave one environment out\" cross-validation, i.e. 3 trials of training on two environments and testing on the third. The baseline model is first trained on data for two environments and then used to generate the n = 50 best plans for both training and testing instructions. As mentioned in Section 3.1, we need to generate many more top parse trees to get 50 distinct formal MR plans. We limit the number of best parse trees to 1,000,000, and even with this high limit, some training examples were left with less than 50 distinct plans. Table 1 : Oracle parse and execution accuracy for single sentence and complete paragraph instructions for the n best parses.",
"cite_spans": [
{
"start": 687,
"end": 709,
"text": "MacMahon et al. (2006)",
"ref_id": "BIBREF20"
},
{
"start": 771,
"end": 772,
"text": "1",
"ref_id": null
},
{
"start": 1164,
"end": 1186,
"text": "Chen and Mooney (2011)",
"ref_id": "BIBREF2"
},
{
"start": 1548,
"end": 1569,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
},
{
"start": 1574,
"end": 1596,
"text": "Chen and Mooney (2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [
{
"start": 641,
"end": 649,
"text": "Figure 3",
"ref_id": null
},
{
"start": 2150,
"end": 2157,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Descended Action Features",
"sec_num": "4.3"
},
{
"text": "plan is then executed using MARCO and its rate of successfully reaching the goal is recorded. Our reranking model is then trained on the training data using the n-best candidate parses. We only retain reranking features that appear (i.e. have a value of 1) at least twice in the training data. Finally, we measure both parse and execution accuracy on the test data. Parse accuracy evaluates how well a system maps novel NL sentences for new environments into correct MR plans (Chen and Mooney, 2011) . It is calculated by comparing the system's MR output to the gold-standard MR. Accuracy is measured using F1, the harmonic mean of precision and recall for individual MR constituents, thereby giving partial credit to approximately correct MRs. We then execute the resulting MR plans in the test environment to see whether they successfully reach the desired destinations. Execution is evaluated both for single sentence and complete paragraph instructions. Successful execution rates are calculated by averaging 10 nondeterministic MARCO executions.",
"cite_spans": [
{
"start": 476,
"end": 499,
"text": "(Chen and Mooney, 2011)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Descended Action Features",
"sec_num": "4.3"
},
{
"text": "As typical in reranking experiments, we first present results for an \"oracle\" that always returns the best result amongst the top-n candidates produced by the baseline system, thereby providing an upper bound on the improvements possible with reranking. Table 1 shows oracle accuracy for both semantic parsing and plan execution for single sentence and complete paragraph instructions for various values of n. For oracle parse accuracy, for each sentence, we pick the parse that gives the highest F1 score. For oracle single-sentence execution accuracy, we pick the parse that gives the highest execution success rate. These singlesentence plans are then concatenated to produce a complete plan for each paragraph instruction in order to measure overall execution accuracy. Since making an error in any of the sentences in an in-struction can easily lead to the wrong final destination, paragraph-level accuracies are always much lower than sentence-level ones. In order to balance oracle accuracy and the computational effort required to produce n distinct plans, we chose n = 50 for the final experiments since oracle performance begins to asymptote at this point.",
"cite_spans": [],
"ref_spans": [
{
"start": 254,
"end": 261,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Oracle results",
"sec_num": null
},
{
"text": "Response-based vs. gold-standard reference weight updates Table 2 presents reranking results for our proposed response-based weight update (Single) for the averaged perceptron (cf. Section 3.1) compared to the typical weight update method using goldstandard parses (Gold) . Since the gold-standard annotation gives the correct MR rather than a parse tree for each sentence, Gold selects as a single reference parse the candidate in the top 50 whose resulting MR is most similar to the gold-standard MR as determined by its parse accuracy. Ge and Mooney (2006) employ a similar approach when reranking semantic parses.",
"cite_spans": [
{
"start": 265,
"end": 271,
"text": "(Gold)",
"ref_id": null
},
{
"start": 539,
"end": 559,
"text": "Ge and Mooney (2006)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [
{
"start": 58,
"end": 65,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Oracle results",
"sec_num": null
},
{
"text": "The results show that our response-based approach (Single) has better execution accuracy than both the baseline and the standard approach using gold-standard parses (Gold) . However, Gold does perform best on parse accuracy since it explicitly focuses on maximizing the accuracy of the resulting MR. In contrast, by focusing discriminative training on optimizing performance of the ultimate end task, our response-based approach actually outperforms the traditional approach on the final task. In addition, it only utilizes feedback that is naturally available for the task, rather than requiring an expert to laboriously annotate each sentence with a gold-standard MR. Even though Gold captures more elements of the gold-standard MRs, it may miss some critical MR components that are crucial to the final navigation task. The overall result is very promising because it demonstrates how reranking can be applied to grounded language learning tasks where gold-standard parses are not readily available. Table 2 : Reranking results comparing our response-based methods using single (Single) or multiple (Multi) pseudo-gold parses to the standard approach using a single gold-standard parse (Gold) . Baseline refers to Kim and Mooney (2012) 's system. Reranking results use all features described in Section 4. \"Single\" means the single-sentence version and \"Para\" means the full paragraph version of the corpus.",
"cite_spans": [
{
"start": 165,
"end": 171,
"text": "(Gold)",
"ref_id": null
},
{
"start": 1189,
"end": 1195,
"text": "(Gold)",
"ref_id": null
},
{
"start": 1217,
"end": 1238,
"text": "Kim and Mooney (2012)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [
{
"start": 1003,
"end": 1010,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Oracle results",
"sec_num": null
},
{
"text": "Weight update with single vs. multiple reference parses Table 2 also shows performance when using multiple reference parse trees to update weights (cf. Section 3.2). Using multiple parses (Multi) clearly performs better for all evaluation metrics, particularly execution. As explained in Section 3.2, the single-best pseudo-gold parse provides weak, ambiguous feedback since it only provides a rough estimate of the response feedback from the execution module. Using a variety of preferable parses to update weights provides a greater amount and variety of weak feedback and therefore leads to a more accurate model. 3",
"cite_spans": [],
"ref_spans": [
{
"start": 56,
"end": 63,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Oracle results",
"sec_num": null
},
{
"text": "Comparison of different feature groups Table 3 compares reranking results using the different feature groups described in Section 4. Compared to the baseline model (Kim and Mooney, 2012) , each of the feature groups Base (base features), Pred (predicate-only and verificationremoved features), and Desc (descended action features) helps improve the performance of plan execution for both single sentence and complete paragraph navigation instructions. Among them, Desc is the most effective group of features. Combinations of the feature groups helps fur- 3 We also tried extending Gold to use multiple reference parses in the same manner, but this actually degraded its performance for all metrics. This indicates that, unlike Multi, parses other than the best one do not have useful information in terms of optimizing normal parse accuracy. Instead, additional parses seem to add noise to the training process in this case. Therefore, updating with multiple parses does not appear to be useful in standard reranking. ther improve the plan execution performance, and reranking using all of the feature groups (All) performs the best, as expected. However, since our model is optimizing plan execution during training, the results for parse accuracy are always worse than the baseline model.",
"cite_spans": [
{
"start": 164,
"end": 186,
"text": "(Kim and Mooney, 2012)",
"ref_id": "BIBREF16"
},
{
"start": 556,
"end": 557,
"text": "3",
"ref_id": null
}
],
"ref_spans": [
{
"start": 39,
"end": 46,
"text": "Table 3",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "Oracle results",
"sec_num": null
},
{
"text": "Discriminative reranking is a common machine learning technique to improve the output of generative models. It has been shown to be effective for various natural language processing tasks including syntactic parsing (Collins, 2000; Collins, 2002b; Collins and Koo, 2005; Charniak and Johnson, 2005; Huang, 2008) , semantic parsing (Lu et al., 2008; Ge and Mooney, 2006) , partof-speech tagging (Collins, 2002a) , semantic role labeling (Toutanova et al., 2005) , named entity recognition (Collins, 2002c) . machine translation (Shen et al., 2004; Fraser and Marcu, 2006) and surface realization in generation (White and Rajkumar, 2009; Konstas and Lapata, 2012) . However, to our knowledge, there has been no previous attempt to apply discriminative reranking to grounded language acquisition, where goldstandard reference parses are not typically available for training reranking models. Our use of response-based training is similar to work on learning semantic parsers from execution output such as the answers to database queries (Clarke et al., 2010; Liang et al., 2011) . Although the demands of grounded language tasks, such as following navigation instructions, are different, it would be interesting to try adapting these alternative approaches to such problems.",
"cite_spans": [
{
"start": 216,
"end": 231,
"text": "(Collins, 2000;",
"ref_id": "BIBREF6"
},
{
"start": 232,
"end": 247,
"text": "Collins, 2002b;",
"ref_id": null
},
{
"start": 248,
"end": 270,
"text": "Collins and Koo, 2005;",
"ref_id": "BIBREF5"
},
{
"start": 271,
"end": 298,
"text": "Charniak and Johnson, 2005;",
"ref_id": "BIBREF1"
},
{
"start": 299,
"end": 311,
"text": "Huang, 2008)",
"ref_id": "BIBREF14"
},
{
"start": 331,
"end": 348,
"text": "(Lu et al., 2008;",
"ref_id": "BIBREF19"
},
{
"start": 349,
"end": 369,
"text": "Ge and Mooney, 2006)",
"ref_id": "BIBREF11"
},
{
"start": 394,
"end": 410,
"text": "(Collins, 2002a)",
"ref_id": "BIBREF7"
},
{
"start": 436,
"end": 460,
"text": "(Toutanova et al., 2005)",
"ref_id": "BIBREF23"
},
{
"start": 488,
"end": 504,
"text": "(Collins, 2002c)",
"ref_id": "BIBREF9"
},
{
"start": 527,
"end": 546,
"text": "(Shen et al., 2004;",
"ref_id": "BIBREF22"
},
{
"start": 547,
"end": 570,
"text": "Fraser and Marcu, 2006)",
"ref_id": "BIBREF10"
},
{
"start": 609,
"end": 635,
"text": "(White and Rajkumar, 2009;",
"ref_id": "BIBREF24"
},
{
"start": 636,
"end": 661,
"text": "Konstas and Lapata, 2012)",
"ref_id": "BIBREF17"
},
{
"start": 1034,
"end": 1055,
"text": "(Clarke et al., 2010;",
"ref_id": "BIBREF4"
},
{
"start": 1056,
"end": 1075,
"text": "Liang et al., 2011)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "6"
},
{
"text": "In the future, we would like to explore the construction of better, more-general reranking features that are less prone to over-fitting. Since typical reranking features rely on the combination and/or modification of nonterminals appearing in parse trees, for the large PCFG's produced for grounded language learning, such features are very sparse and rare. Although the current features provide a significant increase in performance, oracle results imply that an even larger benefit may be achievable.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Future Work",
"sec_num": "7"
},
{
"text": "In addition, employing other reranking methodologies, such as kernel methods (Collins, 2002b) , and forest reranking exploiting a packed forest of exponentially many parse trees (Huang, 2008) , is another area of future work. We also would like to apply our approach to other reranking algorithms such as SVMs (Joachims, 2002) and Max-Ent methods (Charniak and Johnson, 2005) .",
"cite_spans": [
{
"start": 77,
"end": 93,
"text": "(Collins, 2002b)",
"ref_id": null
},
{
"start": 178,
"end": 191,
"text": "(Huang, 2008)",
"ref_id": "BIBREF14"
},
{
"start": 310,
"end": 326,
"text": "(Joachims, 2002)",
"ref_id": "BIBREF15"
},
{
"start": 347,
"end": 375,
"text": "(Charniak and Johnson, 2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Future Work",
"sec_num": "7"
},
{
"text": "In this paper, we have shown how to adapt discriminative reranking to grounded language learning. Since typical grounded language learning problems, such as navigation instruction following, do not provide the gold-standard reference parses required by standard reranking models, we have devised a novel method for using the weaker supervision provided by response feedback (e.g. the execution of inferred navigation plans) when training a perceptron-based reranker. This approach was shown to be very effective compared to the traditional method of using gold-standard parses. In addition, since this response-based supervision is weak and ambiguous, we have also presented a method for using multiple reference parses to perform perceptron weight updates and shown a clear further improvement in end-task performance with this approach.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions",
"sec_num": "8"
},
{
"text": "Data is available at http://www.cs.utexas. edu/users/ml/clamp/navigation/ 2 9.6% of the examples (310 out of total 3237) produced less than 50 distinct MR plans in the evaluation. This was mostly due to exceeding the parse-tree limit and partly because the baseline model failed to parse some NL sentences.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Michael Collins. 2002b. New ranking algorithms for parsing and tagging: Kernels over discrete structures, and the voted perceptron. In Proceedings of",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "We thank anonymous reviewers for their helpful comments to improve this paper. This work was funded by the NSF grant IIS-0712907 and IIS-1016312. Experiments were performed on the Mastodon Cluster, provided by NSF Grant EIA-0303609.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Reducing grounded learning tasks to grammatical inference",
"authors": [
{
"first": "Benjamin",
"middle": [],
"last": "B\u00f6rschinger",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Bevan",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Jones",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Johnson",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "1416--1425",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Benjamin B\u00f6rschinger, Bevan K. Jones, and Mark Johnson. 2011. Reducing grounded learning tasks to grammatical inference. In Proceedings of the Conference on Empirical Methods in Natural Lan- guage Processing, EMNLP '11, pages 1416-1425, Stroudsburg, PA, USA. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Coarseto-fine n-best parsing and maxent discriminative reranking",
"authors": [
{
"first": "Eugene",
"middle": [],
"last": "Charniak",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Johnson",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the 43nd Annual Meeting of the Association for Computational Linguistics (ACL-05)",
"volume": "",
"issue": "",
"pages": "173--180",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Eugene Charniak and Mark Johnson. 2005. Coarse- to-fine n-best parsing and maxent discriminative reranking. In Proceedings of the 43nd Annual Meet- ing of the Association for Computational Linguistics (ACL-05), pages 173-180, Ann Arbor, MI, June.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Learning to interpret natural language navigation instructions from observations",
"authors": [
{
"first": "L",
"middle": [],
"last": "David",
"suffix": ""
},
{
"first": "Raymond",
"middle": [
"J"
],
"last": "Chen",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mooney",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of the 25th AAAI Conference on Artificial Intelligence (AAAI-2011)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "David L. Chen and Raymond J. Mooney. 2011. Learn- ing to interpret natural language navigation instruc- tions from observations. In Proceedings of the 25th AAAI Conference on Artificial Intelligence (AAAI- 2011), San Francisco, CA, USA, August.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Training a multilingual sportscaster: Using perceptual context to learn language",
"authors": [
{
"first": "David",
"middle": [
"L"
],
"last": "Chen",
"suffix": ""
},
{
"first": "Joohyun",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Raymond",
"middle": [
"J"
],
"last": "Mooney",
"suffix": ""
}
],
"year": 2010,
"venue": "Journal of Artificial Intelligence Research",
"volume": "37",
"issue": "",
"pages": "397--435",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "David L. Chen, Joohyun Kim, and Raymond J. Mooney. 2010. Training a multilingual sportscaster: Using perceptual context to learn language. Journal of Artificial Intelligence Research, 37:397-435.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Driving semantic parsing from the world's response",
"authors": [
{
"first": "James",
"middle": [],
"last": "Clarke",
"suffix": ""
},
{
"first": "Dan",
"middle": [],
"last": "Goldwasser",
"suffix": ""
},
{
"first": "Ming-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Dan",
"middle": [
"Roth"
],
"last": "",
"suffix": ""
}
],
"year": 2010,
"venue": "Proceedings of the Fourteenth Conference on Computational Natural Language Learning (CoNLL-2010)",
"volume": "",
"issue": "",
"pages": "18--27",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "James Clarke, Dan Goldwasser, Ming-Wei Chang, and Dan Roth. 2010. Driving semantic parsing from the world's response. In Proceedings of the Four- teenth Conference on Computational Natural Lan- guage Learning (CoNLL-2010), pages 18-27, Upp- sala, Sweden, July. Association for Computational Linguistics.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Discriminative reranking for natural language parsing",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
},
{
"first": "Terry",
"middle": [],
"last": "Koo",
"suffix": ""
}
],
"year": 2005,
"venue": "Computational Linguistics",
"volume": "31",
"issue": "1",
"pages": "25--69",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins and Terry Koo. 2005. Discriminative reranking for natural language parsing. Computa- tional Linguistics, 31(1):25-69.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Discriminative reranking for natural language parsing",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2000,
"venue": "Proceedings of the Seventeenth International Conference on Machine Learning (ICML-2000)",
"volume": "",
"issue": "",
"pages": "175--182",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 2000. Discriminative reranking for natural language parsing. In Proceedings of the Seventeenth International Conference on Machine Learning (ICML-2000), pages 175-182, Stanford, CA, June.",
"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 2002 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 2002a. Discriminative training meth- ods for hidden Markov models: Theory and ex- periments with perceptron algorithms. In Proceed- ings of the 2002 Conference on Empirical Meth- ods in Natural Language Processing (EMNLP-02), Philadelphia, PA, July.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "the 40th Annual Meeting of the Association for Computational Linguistics (ACL-2002)",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "263--270",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "the 40th Annual Meeting of the Association for Com- putational Linguistics (ACL-2002), pages 263-270, Philadelphia, PA, July.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Ranking algorithms for named-entity extraction: Boosting and the voted perceptron",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics (ACL-2002)",
"volume": "",
"issue": "",
"pages": "489--496",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 2002c. Ranking algorithms for named-entity extraction: Boosting and the voted perceptron. In Proceedings of the 40th Annual Meeting of the Association for Computational Lin- guistics (ACL-2002), pages 489-496, Philadelphia, PA.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Semisupervised training for statistical word alignment",
"authors": [
{
"first": "Alexander",
"middle": [],
"last": "Fraser",
"suffix": ""
},
{
"first": "Daniel",
"middle": [],
"last": "Marcu",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics (ACL-06)",
"volume": "",
"issue": "",
"pages": "769--776",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alexander Fraser and Daniel Marcu. 2006. Semi- supervised training for statistical word alignment. In Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Lin- guistics (ACL-06), pages 769-776, Stroudsburg, PA, USA. Association for Computational Linguistics.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Discriminative reranking for semantic parsing",
"authors": [
{
"first": "R",
"middle": [],
"last": "Ge",
"suffix": ""
},
{
"first": "R",
"middle": [
"J"
],
"last": "Mooney",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics (COLING/ACL-06)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R. Ge and R. J. Mooney. 2006. Discriminative reranking for semantic parsing. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Associa- tion for Computational Linguistics (COLING/ACL- 06), Sydney, Australia, July.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "A robot that uses existing vocabulary to infer non-visual word meanings from observation",
"authors": [
{
"first": "Kevin",
"middle": [],
"last": "Gold",
"suffix": ""
},
{
"first": "Brian",
"middle": [],
"last": "Scassellati",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 22nd national conference on Artificial intelligence",
"volume": "1",
"issue": "",
"pages": "883--888",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kevin Gold and Brian Scassellati. 2007. A robot that uses existing vocabulary to infer non-visual word meanings from observation. In Proceedings of the 22nd national conference on Artificial intelligence - Volume 1, AAAI'07, pages 883-888. AAAI Press.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Better kbest parsing",
"authors": [
{
"first": "Liang",
"middle": [],
"last": "Huang",
"suffix": ""
},
{
"first": "David",
"middle": [],
"last": "Chiang",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the Ninth International Workshop on Parsing Technology, Parsing '05",
"volume": "",
"issue": "",
"pages": "53--64",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Liang Huang and David Chiang. 2005. Better k- best parsing. In Proceedings of the Ninth Inter- national Workshop on Parsing Technology, Parsing '05, pages 53-64, Stroudsburg, PA, USA. Associa- tion for Computational Linguistics.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Forest reranking: Discriminative parsing with non-local features",
"authors": [
{
"first": "Liang",
"middle": [],
"last": "Huang",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of ACL-08: HLT",
"volume": "",
"issue": "",
"pages": "586--594",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Liang Huang. 2008. Forest reranking: Discrimina- tive parsing with non-local features. In Proceedings of ACL-08: HLT, pages 586-594, Columbus, Ohio, June. Association for Computational Linguistics.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Optimizing search engines using clickthrough data",
"authors": [
{
"first": "Thorsten",
"middle": [],
"last": "Joachims",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the Eighth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Thorsten Joachims. 2002. Optimizing search en- gines using clickthrough data. In Proceedings of the Eighth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD- 2002), Edmonton, Canada.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Unsupervised PCFG induction for grounded language learning with highly ambiguous supervision",
"authors": [
{
"first": "Joohyun",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Raymond",
"middle": [
"J"
],
"last": "Mooney",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the Conference on Empirical Methods in Natural Language Processing and Natural Language Learning, EMNLP-CoNLL '12",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Joohyun Kim and Raymond J. Mooney. 2012. Un- supervised PCFG induction for grounded language learning with highly ambiguous supervision. In Pro- ceedings of the Conference on Empirical Methods in Natural Language Processing and Natural Lan- guage Learning, EMNLP-CoNLL '12.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Conceptto-text generation via discriminative reranking",
"authors": [
{
"first": "Ioannis",
"middle": [],
"last": "Konstas",
"suffix": ""
},
{
"first": "Mirella",
"middle": [],
"last": "Lapata",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Long Papers",
"volume": "1",
"issue": "",
"pages": "369--378",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ioannis Konstas and Mirella Lapata. 2012. Concept- to-text generation via discriminative reranking. In Proceedings of the 50th Annual Meeting of the Asso- ciation for Computational Linguistics: Long Papers -Volume 1, ACL '12, pages 369-378, Stroudsburg, PA, USA. Association for Computational Linguis- tics.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Learning dependency-based compositional semantics",
"authors": [
{
"first": "Percy",
"middle": [],
"last": "Liang",
"suffix": ""
},
{
"first": "Michael",
"middle": [
"I"
],
"last": "Jordan",
"suffix": ""
},
{
"first": "Dan",
"middle": [],
"last": "Klein",
"suffix": ""
}
],
"year": 2011,
"venue": "Proceedings of ACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Percy Liang, Michael I. Jordan, and Dan Klein. 2011. Learning dependency-based compositional seman- tics. In Proceedings of ACL, Portland, Oregon, June. Association for Computational Linguistics.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "A generative model for parsing natural language to meaning representations",
"authors": [
{
"first": "Wei",
"middle": [],
"last": "Lu",
"suffix": ""
},
{
"first": "Wee",
"middle": [],
"last": "Hwee Tou Ng",
"suffix": ""
},
{
"first": "Luke",
"middle": [
"S"
],
"last": "Sun Lee",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wei Lu, Hwee Tou Ng, Wee Sun Lee, and Luke S. Zettlemoyer. 2008. A generative model for pars- ing natural language to meaning representations. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing (EMNLP- 08), Honolulu, HI, October.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Walk the talk: connecting language, knowledge, and action in route instructions",
"authors": [
{
"first": "Matt",
"middle": [],
"last": "Macmahon",
"suffix": ""
},
{
"first": "Brian",
"middle": [],
"last": "Stankiewicz",
"suffix": ""
},
{
"first": "Benjamin",
"middle": [],
"last": "Kuipers",
"suffix": ""
}
],
"year": 2006,
"venue": "proceedings of the 21st national conference on Artificial intelligence",
"volume": "2",
"issue": "",
"pages": "1475--1482",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Matt MacMahon, Brian Stankiewicz, and Benjamin Kuipers. 2006. Walk the talk: connecting language, knowledge, and action in route instructions. In pro- ceedings of the 21st national conference on Artifi- cial intelligence -Volume 2, AAAI'06, pages 1475- 1482. AAAI Press.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Learning visually grounded words and syntax for a scene description task",
"authors": [
{
"first": "Deb",
"middle": [],
"last": "Roy",
"suffix": ""
}
],
"year": 2002,
"venue": "Computer Speech and Language",
"volume": "16",
"issue": "3",
"pages": "353--385",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Deb Roy. 2002. Learning visually grounded words and syntax for a scene description task. Computer Speech and Language, 16(3):353-385.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Discriminative reranking for machine translation",
"authors": [
{
"first": "Libin",
"middle": [],
"last": "Shen",
"suffix": ""
},
{
"first": "Anoop",
"middle": [],
"last": "Sarkar",
"suffix": ""
},
{
"first": "Franz Josef",
"middle": [],
"last": "Och",
"suffix": ""
}
],
"year": 2004,
"venue": "HLT-NAACL 2004: Main Proceedings",
"volume": "",
"issue": "",
"pages": "177--184",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Libin Shen, Anoop Sarkar, and Franz Josef Och. 2004. Discriminative reranking for machine translation. In Daniel Marcu Susan Dumais and Salim Roukos, ed- itors, HLT-NAACL 2004: Main Proceedings, pages 177-184, Boston, Massachusetts, USA, May 2 - May 7. Association for Computational Linguistics.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "Joint learning improves semantic role labeling",
"authors": [
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
},
{
"first": "Aria",
"middle": [],
"last": "Haghighi",
"suffix": ""
},
{
"first": "Christopher",
"middle": [
"D"
],
"last": "Manning",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the 43nd Annual Meeting of the Association for Computational Linguistics (ACL-05)",
"volume": "",
"issue": "",
"pages": "589--596",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kristina Toutanova, Aria Haghighi, and Christopher D. Manning. 2005. Joint learning improves semantic role labeling. In Proceedings of the 43nd Annual Meeting of the Association for Computational Lin- guistics (ACL-05), pages 589-596, Ann Arbor, MI, June.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Perceptron reranking for CCG realization",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "White",
"suffix": ""
},
{
"first": "Rajakrishnan",
"middle": [],
"last": "Rajkumar",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing",
"volume": "1",
"issue": "",
"pages": "410--419",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael White and Rajakrishnan Rajkumar. 2009. Perceptron reranking for CCG realization. In Pro- ceedings of the 2009 Conference on Empirical Meth- ods in Natural Language Processing: Volume 1 - Volume 1, EMNLP '09, pages 410-419, Strouds- burg, PA, USA. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "On the integration of grounding language and learning objects",
"authors": [
{
"first": "Chen",
"middle": [],
"last": "Yu",
"suffix": ""
},
{
"first": "Dana",
"middle": [
"H"
],
"last": "Ballard",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the Nineteenth National Conference on Artificial Intelligence (AAAI-04)",
"volume": "",
"issue": "",
"pages": "488--493",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chen Yu and Dana H. Ballard. 2004. On the integra- tion of grounding language and learning objects. In Proceedings of the Nineteenth National Conference on Artificial Intelligence (AAAI-04), pages 488-493.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "Sample virtual world and instruction."
},
"FIGREF1": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "Simplified parse for the sentence \"Turn left and find the sofa then turn around the corner\" for Kim and Mooney's model. Nonterminals show the MR graph, where additional nonterminals for generating NL words are omitted."
},
"FIGREF2": {
"uris": null,
"type_str": "figure",
"num": null,
"text": ", f (Turn(LEFT)) = 1 since it descends into L 2 and L 4 . b) Descended Action Unigram. Same as Descended Action, but also includes the current nonterminal: f (Turn(LEFT)|L 1 ) = 1. c) Grandparent Descended Action Unigram. Same as Descended Action Unigram, but additionally includes the parent nonterminal as well as the current one: f (Turn(LEFT)|L 2 , L 1 ) = 1.d) Long-range Descended Action Unigram. Indicates whether a given atomic action in a nonterminal descends to a child nonterminal and this child generates a given NL word below it: f (Turn(LEFT) ; left)"
},
"TABREF0": {
"content": "<table/>",
"html": null,
"type_str": "table",
"num": null,
"text": "Algorithm 1 AVERAGED PERCEPTRON TRAIN-ING WITH RESPONSE-BASED UPDATEInput: A set of training examples (e i , y * i ), where e i is a NL sentence and y * i = arg max y\u2208GEN(e i ) EXEC(y) Output: The parameter vectorW , averaged over all iterations 1...T 1: procedure PERCEPTRON"
},
"TABREF1": {
"content": "<table><tr><td>n</td><td/><td>1</td><td>2</td><td>5</td><td>10</td><td>25</td><td>50</td></tr><tr><td>Parse Accuracy</td><td>F1</td><td>74.81</td><td/><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>2 Each candidate</td></tr></table>",
"html": null,
"type_str": "table",
"num": null,
"text": "79.08 82.78 85.32 87.52 88.62 Plan Execution Single-sentence 57.22 63.86 70.93 76.41 83.59 87.02 Paragraph 20.17 28.08 35.34 40.64 48.69 53.66"
},
"TABREF4": {
"content": "<table><tr><td>Section 4.1), Pred refers to predicate-only fea-</td></tr><tr><td>tures and also includes features based on remov-</td></tr><tr><td>ing interleaving verification steps (cf. Section 4.2),</td></tr><tr><td>Desc refers to descended action features (cf. Sec-</td></tr><tr><td>tion 4.3). All refers to all the features including</td></tr><tr><td>Base, Pred, and Desc. All results use weight</td></tr><tr><td>update with multiple reference parses (cf. Sec-</td></tr><tr><td>tion 3.2).</td></tr></table>",
"html": null,
"type_str": "table",
"num": null,
"text": "Reranking results comparing different sets of features. Base refers to base features (cf."
}
}
}
} |