File size: 89,394 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 | {
"paper_id": "P06-1029",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T09:26:35.390594Z"
},
"title": "Approximation Lasso Methods for Language Modeling",
"authors": [
{
"first": "Jianfeng",
"middle": [],
"last": "Gao",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Microsoft Research One Microsoft Way Redmond",
"location": {
"postCode": "98052",
"region": "WA",
"country": "USA"
}
},
"email": "jfgao@microsoft.com"
},
{
"first": "Hisami",
"middle": [],
"last": "Suzuki",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Microsoft Research",
"location": {
"addrLine": "One Microsoft Way",
"postCode": "98052",
"settlement": "Redmond",
"region": "WA",
"country": "USA"
}
},
"email": "hisamis@microsoft.com"
},
{
"first": "Bin",
"middle": [],
"last": "Yu",
"suffix": "",
"affiliation": {},
"email": "binyu@stat.berkeley.edu"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Lasso is a regularization method for parameter estimation in linear models. It optimizes the model parameters with respect to a loss function subject to model complexities. This paper explores the use of lasso for statistical language modeling for text input. Owing to the very large number of parameters, directly optimizing the penalized lasso loss function is impossible. Therefore, we investigate two approximation methods, the boosted lasso (BLasso) and the forward stagewise linear regression (FSLR). Both methods, when used with the exponential loss function, bear strong resemblance to the boosting algorithm which has been used as a discriminative training method for language modeling. Evaluations on the task of Japanese text input show that BLasso is able to produce the best approximation to the lasso solution, and leads to a significant improvement, in terms of character error rate, over boosting and the traditional maximum likelihood estimation.",
"pdf_parse": {
"paper_id": "P06-1029",
"_pdf_hash": "",
"abstract": [
{
"text": "Lasso is a regularization method for parameter estimation in linear models. It optimizes the model parameters with respect to a loss function subject to model complexities. This paper explores the use of lasso for statistical language modeling for text input. Owing to the very large number of parameters, directly optimizing the penalized lasso loss function is impossible. Therefore, we investigate two approximation methods, the boosted lasso (BLasso) and the forward stagewise linear regression (FSLR). Both methods, when used with the exponential loss function, bear strong resemblance to the boosting algorithm which has been used as a discriminative training method for language modeling. Evaluations on the task of Japanese text input show that BLasso is able to produce the best approximation to the lasso solution, and leads to a significant improvement, in terms of character error rate, over boosting and the traditional maximum likelihood estimation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Language modeling (LM) is fundamental to a wide range of applications. Recently, it has been shown that a linear model estimated using discriminative training methods, such as the boosting and perceptron algorithms, outperforms significantly a traditional word trigram model trained using maximum likelihood estimation (MLE) on several tasks such as speech recognition and Asian language text input (Bacchiani et al. 2004; Roark et al. 2004; Gao et al. 2005; Suzuki and Gao 2005) .",
"cite_spans": [
{
"start": 399,
"end": 422,
"text": "(Bacchiani et al. 2004;",
"ref_id": "BIBREF0"
},
{
"start": 423,
"end": 441,
"text": "Roark et al. 2004;",
"ref_id": "BIBREF10"
},
{
"start": 442,
"end": 458,
"text": "Gao et al. 2005;",
"ref_id": "BIBREF7"
},
{
"start": 459,
"end": 479,
"text": "Suzuki and Gao 2005)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The success of discriminative training methods is largely due to fact that unlike the traditional approach (e.g., MLE) that maximizes the function (e.g., likelihood of training data) that is loosely associated with error rate, discriminative training methods aim to directly minimize the error rate on training data even if they reduce the likelihood. However, given a finite set of training samples, discriminative training methods could lead to an arbitrary complex model for the purpose of achieving zero training error. It is well-known that complex models exhibit high variance and perform poorly on unseen data. Therefore some regularization methods have to be used to control the complexity of the model.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Lasso is a regularization method for parameter estimation in linear models. It optimizes the model parameters with respect to a loss function subject to model complexities. The basic idea of lasso is originally proposed by Tibshirani (1996) . Recently, there have been several implementations and experiments of lasso on multi-class classification tasks where only a small number of features need to be handled and the lasso solution can be directly computed via numerical methods. To our knowledge, this paper presents the first empirical study of lasso for a realistic, large scale task: LM for Asian language text input. Because the task utilizes millions of features and training samples, directly optimizing the penalized lasso loss function is impossible. Therefore, two approximation methods, the boosted lasso (BLasso, Zhao and Yu 2004) and the forward stagewise linear regression (FSLR, Hastie et al. 2001) , are investigated. Both methods, when used with the exponential loss function, bear strong resemblance to the boosting algorithm which has been used as a discriminative training method for LM. Evaluations on the task of Japanese text input show that BLasso is able to produce the best approximation to the lasso solution, and leads to a significant improvement, in terms of character error rate, over the boosting algorithm and the traditional MLE.",
"cite_spans": [
{
"start": 223,
"end": 240,
"text": "Tibshirani (1996)",
"ref_id": "BIBREF13"
},
{
"start": 827,
"end": 844,
"text": "Zhao and Yu 2004)",
"ref_id": "BIBREF15"
},
{
"start": 889,
"end": 915,
"text": "(FSLR, Hastie et al. 2001)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper studies LM on the application of Asian language (e.g. Chinese or Japanese) text input, a standard method of inputting Chinese or Japanese text by converting the input phonetic symbols into the appropriate word string. In this paper we call the task IME, which stands for input method editor, based on the name of the commonly used Windows-based application.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "Performance on IME is measured in terms of the character error rate (CER), which is the number of characters wrongly converted from the phonetic string divided by the number of characters in the correct transcript.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "Similar to speech recognition, IME is viewed as a Bayes decision problem. Let A be the input phonetic string. An IME system's task is to choose the most likely word string W * among those candidates that could be converted from A:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": ") | ( ) ( max arg ) | ( max arg (A) ) ( * W A P W P A W P W W A W GEN GEN \u2208 \u2208 = = (1)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "where GEN(A) denotes the candidate set given A. Unlike speech recognition, however, there is no acoustic ambiguity as the phonetic string is inputted by users. Moreover, we can assume a unique mapping from W and A in IME as words have unique readings, i.e. P(A|W) = 1. So the decision of Equation (1) depends solely upon P(W), making IME an ideal evaluation test bed for LM. In this study, the LM task for IME is formulated under the framework of linear models (e.g., Duda et al. 2001) . We use the following notation, adapted from Collins and Koo (2005) :",
"cite_spans": [
{
"start": 468,
"end": 485,
"text": "Duda et al. 2001)",
"ref_id": "BIBREF2"
},
{
"start": 532,
"end": 554,
"text": "Collins and Koo (2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "\u2022 Training data is a set of example input/output pairs. In LM for IME, training samples are represented as {Ai, Wi R }, for i = 1\u2026M, where each Ai is an input phonetic string and Wi R is the reference transcript of Ai.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "\u2022 We assume some way of generating a set of candidate word strings given A, denoted by GEN(A). In our experiments, GEN(A) consists of top n word strings converted from A using a baseline IME system that uses only a word trigram model. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "\u2211 = = D d d d W f \u03bb 0 ) ( .",
"eq_num": "(2)"
}
],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "The decision rule of Equation (1) is rewritten as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": ") , ( max arg ) , ( (A) * \u03bb \u03bb GEN W Score A W W\u2208 = .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "(3)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "Equation 3views IME as a ranking problem, where the model gives the ranking score, not probabilities. We therefore do not evaluate the model via perplexity. Now, assume that we can measure the number of conversion errors in W by comparing it with a reference transcript W R using an error function Er(W R ,W), which is the string edit distance function in our case. We call the sum of error counts over the training samples sample risk. Our goal then is to search for the best parameter set \u03bb which minimizes the sample risk, as in Equation 4:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "\u2211 = = M i i i R i def MSR A W W ... 1 * )) , ( , Er( min arg \u03bb \u03bb \u03bb .",
"eq_num": "(4)"
}
],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "However, (4) cannot be optimized easily since Er(.) is a piecewise constant (or step) function of \u03bb and its gradient is undefined. Therefore, discriminative methods apply different approaches that optimize it approximately. The boosting algorithm described below is one of such approaches.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "LM Task and Problem Definition",
"sec_num": "2"
},
{
"text": "This section gives a brief review of the boosting algorithm, following the description of some recent work (e.g., Schapire and Singer 1999; Collins and Koo 2005) . The boosting algorithm uses an exponential loss function (ExpLoss) to approximate the sample risk in Equation (4). We define the margin of the pair (W R , W) with respect to the model \u03bb as",
"cite_spans": [
{
"start": 114,
"end": 139,
"text": "Schapire and Singer 1999;",
"ref_id": "BIBREF11"
},
{
"start": 140,
"end": 161,
"text": "Collins and Koo 2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": ") , ( ) , ( ) , ( \u03bb \u03bb W Score W Score W W M R R \u2212 =",
"eq_num": "(5)"
}
],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Then, ExpLoss is defined as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "\u2211 \u2211 = \u2208 \u2212 = M i A W i R i i i W W M ... 1 ) ( )) , ( exp( ) ExpLoss( GEN \u03bb (6)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Notice that ExpLoss is convex so there is no problem with local minima when optimizing it. It is shown in Freund et al. (1998) and Collins and Koo (2005) that there exist gradient search procedures that converge to the right solution. Figure 1 summarizes the boosting algorithm we used. After initialization, Steps 2 and 3 are 1 Set \u03bb0 = argmin\u03bb0ExpLoss(\u03bb); and \u03bbd = 0 for d=1\u2026D 2 Select a feature fk* which has largest estimated impact on reducing ExpLoss of Eq. (6) 3 Update \u03bbk* \u03bbk* + \u03b4*, and return to Step 2 repeated N times; at each iteration, a feature is chosen and its weight is updated as follows. First, we define Upd(\u03bb, k, \u03b4) as an updated model, with the same parameter values as \u03bb with the exception of \u03bbk, which is incremented by \u03b4",
"cite_spans": [
{
"start": 106,
"end": 126,
"text": "Freund et al. (1998)",
"ref_id": "BIBREF4"
},
{
"start": 131,
"end": 153,
"text": "Collins and Koo (2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [
{
"start": 235,
"end": 243,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "} ,..., ,..., , { ) , , Upd( 1 0 D k k \u03bb \u03b4 \u03bb \u03bb \u03bb \u03b4 + = \u03bb",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Then, Steps 2 and 3 in Figure 1 can be rewritten as Equations 7and 8",
"cite_spans": [],
"ref_spans": [
{
"start": 23,
"end": 31,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": ", respectively. )) , , d( ExpLoss(Up min arg *) *, ( , \u03b4 \u03b4 \u03b4 k k k \u03bb = (7) *) *, , Upd( 1 \u03b4 k t t \u2212 = \u03bb \u03bb (8)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "The boosting algorithm can be too greedy: Each iteration usually reduces the ExpLoss(.) on training data, so for the number of iterations large enough this loss can be made arbitrarily small. However, fitting training data too well eventually leads to overfiting, which degrades the performance on unseen test data (even though in boosting overfitting can happen very slowly).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Shrinkage is a simple approach to dealing with the overfitting problem. It scales the incremental step \u03b4 by a small constant \u03bd, \u03bd \u2208 (0, 1). Thus, the update of Equation 8",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "with shrinkage is *) *, , Upd( 1 \u03bd\u03b4 k t t \u2212 = \u03bb \u03bb (9)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Empirically, it has been found that smaller values of \u03bd lead to smaller numbers of test errors.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosting",
"sec_num": "3"
},
{
"text": "Lasso is a regularization method for estimation in linear models (Tibshirani 1996) . It regularizes or shrinks a fitted model through an L1 penalty or constraint.",
"cite_spans": [
{
"start": 65,
"end": 82,
"text": "(Tibshirani 1996)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "Let T(\u03bb) denote the L1 penalty of the model, i.e., T(\u03bb) = \u2211d = 0\u2026D|\u03bbd|. We then optimize the model \u03bb so as to minimize a regularized loss function on training data, called lasso loss defined as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": ") ( ) ExpLoss( ) , LassoLoss( \u03bb \u03bb \u03bb T \u03b1 \u03b1 + = (10)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "where T(\u03bb) generally penalizes larger models (or complex models), and the parameter \u03b1 controls the amount of regularization applied to the estimate. Setting \u03b1 = 0 reverses the LassoLoss to the unregularized ExpLoss; as \u03b1 increases, the model coefficients all shrink, each ultimately becoming zero. In practice, \u03b1 should be adaptively chosen to minimize an estimate of expected loss, e.g., \u03b1 decreases with the increase of the number of iterations.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "Computation of the solution to the lasso problem has been studied for special loss functions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "For least square regression, there is a fast algorithm LARS to find the whole lasso path for different \u03b1' s (Obsborn et al. 2000a; 2000b; Efron et al. 2004) ; for 1-norm SVM, it can be transformed into a linear programming problem with a fast algorithm similar to LARS (Zhu et al. 2003) . However, the solution to the lasso problem for a general convex loss function and an adaptive \u03b1 remains open. More importantly for our purposes, directly minimizing lasso function of Equation 10with respect to \u03bb is not possible when a very large number of model parameters are employed, as in our task of LM for IME. Therefore we investigate below two methods that closely approximate the effect of the lasso, and are very similar to the boosting algorithm.",
"cite_spans": [
{
"start": 108,
"end": 130,
"text": "(Obsborn et al. 2000a;",
"ref_id": null
},
{
"start": 131,
"end": 137,
"text": "2000b;",
"ref_id": "BIBREF9"
},
{
"start": 138,
"end": 156,
"text": "Efron et al. 2004)",
"ref_id": "BIBREF3"
},
{
"start": 269,
"end": 286,
"text": "(Zhu et al. 2003)",
"ref_id": "BIBREF16"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "It is also worth noting the difference between L1 and L2 penalty. The classical Ridge Regression setting uses an L2 penalty in Equation 10i.e., T(\u03bb) = \u2211d = 0\u2026D(\u03bbd) 2 , which is much easier to minimize (for least square loss but not for Ex-pLoss). However, recent research (Donoho et al. 1995) shows that the L1 penalty is better suited for sparse situations, where there are only a small number of features with nonzero weights among all candidate features. We find that our task is indeed a sparse situation: among 860,000 features, in the resulting linear model only around 5,000 features have nonzero weights. We then focus on the L1 penalty. We leave the empirical comparison of the L1 and L2 penalty on the LM task to future work.",
"cite_spans": [
{
"start": 272,
"end": 292,
"text": "(Donoho et al. 1995)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Lasso",
"sec_num": "4"
},
{
"text": "The first approximation method we used is FSLR, described in (Algorithm 10.4, Hastie et al. 2001) , where Steps 2 and 3 in Figure 1 are performed according to Equations (7) and (11), respectively.",
"cite_spans": [
{
"start": 78,
"end": 97,
"text": "Hastie et al. 2001)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [
{
"start": 123,
"end": 131,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Forward Stagewise Linear Regression (FSLR)",
"sec_num": "4.1"
},
{
"text": ")) , , d( ExpLoss(Up min arg *) *, ( , \u03b4 \u03b4 \u03b4 k k k \u03bb = (7) *)) sign( *, , Upd( 1 \u03b4 \u03b5 \u00d7 = \u2212 k t t \u03bb \u03bb (11)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Forward Stagewise Linear Regression (FSLR)",
"sec_num": "4.1"
},
{
"text": "Notice that FSLR is very similar to the boosting algorithm with shrinkage in that at each step, the feature fk* that has largest estimated impact on reducing ExpLoss is selected. The only difference is that FSLR updates the weight of fk* by a small fixed step size \u03b5. By taking such small steps, FSLR imposes some implicit regularization, and can closely approximate the effect of the lasso in a local sense (Hastie et al. 2001) . Empirically, we find that the performance of the boosting algorithm with shrinkage closely resembles that of FSLR, with the learning rate parameter \u03bd corresponding to \u03b5.",
"cite_spans": [
{
"start": 408,
"end": 428,
"text": "(Hastie et al. 2001)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Forward Stagewise Linear Regression (FSLR)",
"sec_num": "4.1"
},
{
"text": "The second method we used is a modified version of the BLasso algorithm described in Zhao and Yu (2004) . There are two major differences between BLasso and FSLR. At each iteration, BLasso can take either a forward step or a backward step. Similar to the boosting algorithm and FSLR, at each forward step, a feature is selected and its weight is updated according to Equations (12) and (13).",
"cite_spans": [
{
"start": 85,
"end": 103,
"text": "Zhao and Yu (2004)",
"ref_id": "BIBREF15"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": ")) , , d( ExpLoss(Up min arg *) *, ( , \u03b4 \u03b4 \u03b5 \u03b4 k k k \u03bb \u00b1 = = (12) *)) sign( *, , Upd( 1 \u03b4 \u03b5 \u00d7 = \u2212 k t t \u03bb \u03bb (13)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "However, there is an important difference between Equations (12) and (7). In the boosting algorithm with shrinkage and FSLR, as shown in Equation 7, a feature is selected by its impact on reducing the loss with its optimal update \u03b4 * . In contract, in BLasso, as shown in Equation 12, the optimization over \u03b4 is removed, and for each feature, its loss is calculated with an update of either +\u03b5 or -\u03b5, i.e., the grid search is used for feature selection. We will show later that this seemingly trivial difference brings a significant improvement.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "The backward step is unique to BLasso. In each iteration, a feature is selected and its weight is updated backward if and only if it leads to a decrease of the lasso loss, as shown in Equations (14) and (15):",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": ") ) sign( , , d( ExpLoss(Up min arg * 0 , \u03b5 \u03bb \u03bb \u00d7 \u2212 = \u2260 k k k k k \u03bb (14) ) ) sign( *, , Upd( * 1 \u03b5 \u03bb \u00d7 \u2212 = \u2212 k t t k \u03bb \u03bb \u03b8 \u03b1 \u03b1 > \u2212 \u2212 \u2212 ) , LassoLoss( ) , LassoLoss( if 1 1 t t t t \u03bb \u03bb (15)",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "where \u03b8 is a tolerance parameter. Figure 2 summarizes the BLasso algorithm we used. After initialization, Steps 4 and 5 are repeated N times; at each iteration, a feature is chosen and its weight is updated either backward or forward by a fixed amount \u03b5. Notice that the value of \u03b1 is adaptively chosen according to the reduction of ExpLoss during training. The algorithm starts with a large initial \u03b1, and then at each forward step the value of \u03b1 decreases until the ExpLoss stops decreasing. This is intuitively desirable: It is expected that most highly effective features are selected in early stages of training, so the reduction of ExpLoss at each step in early stages are more substantial than in later stages. These early steps coincide with the boosting steps most of the time. In other words, the effect of backward steps is more visible at later stages.",
"cite_spans": [],
"ref_spans": [
{
"start": 34,
"end": 42,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "Our implementation of BLasso differs slightly from the original algorithm described in Zhao and Yu (2004) . Firstly, because the value of the base feature f0 is the log probability (assigned by a word trigram model) and has a different range from that of other features as in Equation 2, \u03bb0 is set to optimize ExpLoss in the initialization step (Step 1 in Figure 2 ) and remains fixed during training. As suggested by Collins and Koo (2005) , this ensures that the contribution of the log-likelihood feature f0 is well-calibrated with respect to ExpLoss. Secondly, when updating a feature weight, if the size of the optimal update step (computed via Equation 7) is smaller than \u03b5, we use the optimal step to update the feature.",
"cite_spans": [
{
"start": 87,
"end": 105,
"text": "Zhao and Yu (2004)",
"ref_id": "BIBREF15"
},
{
"start": 418,
"end": 440,
"text": "Collins and Koo (2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [
{
"start": 356,
"end": 364,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "Therefore, in our implementation BLasso does not always take a fixed step; it may take steps whose size is smaller than \u03b5. In our initial experiments we found that both changes (also used in our implementations of boosting and FSLR) were crucial to the performance of the methods.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Boosted Lasso (BLasso)",
"sec_num": "4.2"
},
{
"text": "Initialize \u03bb 0 : set \u03bb0 = argmin\u03bb0ExpLoss(\u03bb), and \u03bbd = 0 for d=1\u2026D. 2 Take a forward step according to Eq. 12and 13, and the updated model is denoted by \u03bb 1 3",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1",
"sec_num": null
},
{
"text": "Initialize \u03b1 = (ExpLoss(\u03bb 0 )-ExpLoss(\u03bb 1 ))/\u03b5 4",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1",
"sec_num": null
},
{
"text": "Take a backward step if and only if it leads to a decrease of LassoLoss according to Eq. 14and 15, where \u03b8 = 0; otherwise 5 Take a forward step according to Eq. (12) and (13);",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1",
"sec_num": null
},
{
"text": "update \u03b1 = min(\u03b1, (ExpLoss(\u03bb t-1 )-ExpLoss(\u03bb t ))/\u03b5 ); and return to Step 4.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1",
"sec_num": null
},
{
"text": "The BLasso algorithm (Zhao and Yu 2004) provides theoretical justifications for BLasso. It has been proved that (1) it guarantees that it is safe for BLasso to start with an initial \u03b1 which is the largest \u03b1 that would allow an \u03b5 step away from 0 (i.e., larger \u03b1's correspond to T(\u03bb)=0); (2) for each value of \u03b1, BLasso performs coordinate descent (i.e., reduces Ex-pLoss by updating the weight of a feature) until there is no descent step; and (3) for each step where the value of \u03b1 decreases, it guarantees that the lasso loss is reduced. As a result, it can be proved that for a finite number of features and \u03b8 = 0, the BLasso algorithm shown in Figure 2 converges to the lasso solution when \u03b5 0.",
"cite_spans": [
{
"start": 21,
"end": 39,
"text": "(Zhao and Yu 2004)",
"ref_id": "BIBREF15"
}
],
"ref_spans": [
{
"start": 648,
"end": 656,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Figure 2:",
"sec_num": null
},
{
"text": "We evaluated the training methods described above in the so-called cross-domain language model adaptation paradigm, where we adapt a model trained on one domain (which we call the background domain) to a different domain (adaptation domain), for which only a small amount of training data is available.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Settings",
"sec_num": "5.1"
},
{
"text": "The data sets we used in our experiments came from five distinct sources of text. A 36-million-word Nikkei Newspaper corpus was used as the background domain, on which the word trigram model was trained. We used four adaptation domains: Yomiuri (newspaper corpus), TuneUp (balanced corpus containing newspapers and other sources of text), Encarta (encyclopedia) and Shincho (collection of novels). All corpora have been pre-word-segmented using a lexicon containing 167,107 entries. For each of the four domains, we created training data consisting of 72K sentences (0.9M~1.7M words) and test data of 5K sentences (65K~120K words) from each adaptation domain. The first 800 and 8,000 sentences of each adaptation training data were also used to show how different sizes of training data affected the performances of various adaptation methods. Another 5K-sentence subset was used as held-out data for each domain.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Settings",
"sec_num": "5.1"
},
{
"text": "We created the training samples for discriminative learning as follows. For each phonetic string A in adaptation training data, we produced a lattice of candidate word strings W using the baseline system described in (Gao et al. 2002) , which uses a word trigram model trained via MLE on the Nikkei Newspaper corpus. For efficiency, we kept only the best 20 hypotheses in its candidate conversion set GEN(A) for each training sample for discriminative training. The oracle best hypothesis, which gives the minimum number of errors, was used as the reference transcript of A.",
"cite_spans": [
{
"start": 217,
"end": 234,
"text": "(Gao et al. 2002)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Settings",
"sec_num": "5.1"
},
{
"text": "We used unigrams and bigrams that occurred more than once in the training set as features in the linear model of Equation (2). The total number of candidate features we used was around 860,000. Table 1 summarizes the results of various model training (adaptation) methods in terms of CER (%) and CER reduction (in parentheses) over comparing models. In the first column, the numbers in parentheses next to the domain name indicates the number of training sentences used for adaptation.",
"cite_spans": [],
"ref_spans": [
{
"start": 194,
"end": 201,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Settings",
"sec_num": "5.1"
},
{
"text": "Baseline, with results shown in Column 3, is the word trigram model. As expected, the CER correlates very well the similarity between the background domain and the adaptation domain, where domain similarity is measured in terms of cross entropy (Yuan et al. 2005) as shown in Column 2.",
"cite_spans": [
{
"start": 245,
"end": 263,
"text": "(Yuan et al. 2005)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "MAP (maximum a posteriori), with results shown in Column 4, is a traditional LM adaptation method where the parameters of the background model are adjusted in such a way that maximizes the likelihood of the adaptation data. Our implementation takes the form of linear interpolation as described in Bacchiani et al. (2004) : P(wi|h) = \u03bbPb(wi|h) + (1-\u03bb)Pa(wi|h), where Pb is the probability of the background model, Pa is the probability trained on adaptation data using MLE and the history h corresponds to two preceding words (i.e. Pb and Pa are trigram probabilities). \u03bb is the interpolation weight optimized on held-out data.",
"cite_spans": [
{
"start": 298,
"end": 321,
"text": "Bacchiani et al. (2004)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "Boosting, with results shown in Column 5, is the algorithm described in Figure 1 . In our implementation, we use the shrinkage method suggested by Schapire and Singer (1999) and Collins and Koo (2005) . At each iteration, we used the following update for the kth feature",
"cite_spans": [
{
"start": 147,
"end": 173,
"text": "Schapire and Singer (1999)",
"ref_id": "BIBREF11"
},
{
"start": 178,
"end": 200,
"text": "Collins and Koo (2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [
{
"start": 72,
"end": 80,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "Z C Z C k k k \u03b5 \u03b5 \u03b4 + + = + _ log 2 1",
"eq_num": "(16)"
}
],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "where Ck + is a value increasing exponentially with the sum of margins of (W R , W) pairs over the set where fk is seen in W R but not in W; Ckis the value related to the sum of margins over the set where fk is seen in W but not in W R . \u03b5 is a smoothing factor (whose value is optimized on held-out data) and Z is a normalization constant (whose value is the ExpLoss(.) of training data according to the current model). We see that \u03b5Z in Equation (16) plays the same role as \u03bd in Equation (9). BLasso, with results shown in Column 6, is the algorithm described in Figure 2 . We find that the performance of BLasso is not very sensitive to the selection of the step size \u03b5 across training sets of different domains and sizes. Although small \u03b5 is preferred in theory as discussed earlier, it would lead to a very slow convergence. Therefore, in our experiments, we always use a large step (\u03b5 = 0.5) and use the so-called early stopping strategy, i.e., the number of iterations before stopping is optimized on held-out data.",
"cite_spans": [],
"ref_spans": [
{
"start": 565,
"end": 573,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "In the task of LM for IME, there are millions of features and training samples, forming an extremely large and sparse matrix. We therefore applied the techniques described in Collins and Koo (2005) to speed up the training procedure. The resulting algorithms run in around 15 and 30 minutes respectively for Boosting and BLasso to converge on an XEON\u2122 MP 1.90GHz machine when training on an 8K-sentnece training set.",
"cite_spans": [
{
"start": 175,
"end": 197,
"text": "Collins and Koo (2005)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "The results in Table 1 give rise to several observations. First of all, both discriminative training methods (i.e., Boosting and BLasso) outperform MAP substantially. The improvement margins are larger when the background and adaptation domains are more similar. The phenomenon is attributed to the underlying difference between the two adaptation methods: MAP aims to improve the likelihood of a distribution, so if the adaptation domain is very similar to the background domain, the difference between the two underlying distributions is so small that MAP cannot adjust the model effectively. Discriminative methods, on the other hand, do not have this limitation for they aim to reduce errors directly. Secondly, BLasso outperforms Boosting significantly (p-value < 0.01) on all test sets. The improvement margins vary with the training sets of different domains and sizes. In general, in cases where the adaptation domain is less similar to the background domain and larger training set is used, the improvement of BLasso is more visible.",
"cite_spans": [],
"ref_spans": [
{
"start": 15,
"end": 22,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "Note that the CER results of FSLR are not included in Table 1 because it achieves very similar results to the boosting algorithm with shrinkage if the controlling parameters of both algorithms are optimized via cross-validation. We shall discuss their difference in the next section.",
"cite_spans": [],
"ref_spans": [
{
"start": 54,
"end": 61,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Main Results",
"sec_num": "5.2"
},
{
"text": "This section investigates what components of BLasso bring the improvement over Boosting. Comparing the algorithms in Figures 1 and 2 , we notice three differences between BLasso and Boosting: (i) the use of backward steps in BLasso; (ii) BLasso uses the grid search (fixed step size) for feature selection in Equation 12while Boosting uses the continuous search (optimal step size) in Equation 7; and (iii) BLasso uses a fixed step size for feature update in Equation 13while Boosting uses an optimal step size in Equation (8). We then investigate these differences in turn.",
"cite_spans": [],
"ref_spans": [
{
"start": 117,
"end": 132,
"text": "Figures 1 and 2",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "To study the impact of backward steps, we compared BLasso with the boosting algorithm with a fixed step search and a fixed step update, henceforth referred to as F-Boosting. F-Boosting was implemented as Figure 2 , by setting a large value to \u03b8 in Equation (15), i.e., \u03b8 = 10 3 , to prohibit backward steps. We find that although the training error curves of BLasso and F-Boosting are almost identical, the T(\u03bb) curves grow apart with iterations, as shown in Figure 3 . The results show that with backward steps, BLasso achieves a better approximation to the true lasso solution:",
"cite_spans": [],
"ref_spans": [
{
"start": 204,
"end": 212,
"text": "Figure 2",
"ref_id": null
},
{
"start": 459,
"end": 467,
"text": "Figure 3",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "It leads to a model with similar training errors but less complex (in terms of L1 penalty). In our experiments we find that the benefit of using backward steps is only visible in later iterations when BLasso's backward steps kick in. A typical example is shown in Figure 4 . The early steps fit to highly effective features and in these steps BLasso and F-Boosting agree. For later steps, fine-tuning of features is required. BLasso with backward steps provides a better mechanism than F-Boosting to revise the previously chosen features to accommodate this fine level of tuning. Consequently we observe the superior performance of BLasso at later stages as shown in our experiments.",
"cite_spans": [],
"ref_spans": [
{
"start": 264,
"end": 272,
"text": "Figure 4",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "As well-known in linear regression models, when there are many strongly correlated features, model parameters can be poorly estimated and exhibit high variance. By imposing a model size constraint, as in lasso, this phenomenon is alleviated. Therefore, we speculate that a better approximation to lasso, as BLasso with backward steps, would be superior in eliminating the negative effect of strongly correlated features in model estimation. To verify our speculation, we performed the following experiments. For each training set, in addition to word unigram and bigram features, we introduced a new type of features called headword bigram.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "As described in Gao et al. (2002) , headwords are defined as the content words of the sentence. Therefore, headword bigrams constitute a special type of skipping bigrams which can capture dependency between two words that may not be adjacent. In reality, a large portion of headword bigrams are identical to word bigrams, as two headwords can occur next to each other in text. In the adaptation test data we used, we find that headword bigram features are for the most part either completely overlapping with the word bigram features (i.e., all instances of headword bigrams also count as word bigrams) or not overlapping at all (i.e., a headword bigram feature is not observed as a word bigram feature) -less than 20% of headword bigram features displayed a variable degree of overlap with word bigram features. In our data, the rate of completely overlapping features is 25% to 47% depending on the adaptation domain. From this, we can say that the headword bigram features show moderate to high degree of correlation with the word bigram features.",
"cite_spans": [
{
"start": 16,
"end": 33,
"text": "Gao et al. (2002)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "We then used BLasso and F-Boosting to train the linear language models including both word bigram and headword bigram features. We find that although the CER reduction by adding headword features is overall very small, the difference between the two versions of BLasso is more visible in all four test sets. Comparing Figure 4 , it can be seen that BLasso with backward steps outperforms the one without backward steps in much earlier stages of training with a larger margin. For example, on Encarta data sets, BLasso outperforms F-Boosting after around 18,000 iterations with headword features (Figure 7) , as opposed to 25,000 iterations without headword features (Figure 4) . The results seem to corroborate our speculation that BLasso is more robust in the presence of highly correlated features.",
"cite_spans": [],
"ref_spans": [
{
"start": 318,
"end": 326,
"text": "Figure 4",
"ref_id": "FIGREF2"
},
{
"start": 595,
"end": 605,
"text": "(Figure 7)",
"ref_id": "FIGREF5"
},
{
"start": 666,
"end": 676,
"text": "(Figure 4)",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "To investigate the impact of using the grid search (fixed step size) versus the continuous search (optimal step size) for feature selection, we compared F-Boosting with FSLR since they differs only in their search methods for feature selection. As shown in Figures 5 to 8 , although FSLR is robust in that its test errors do not increase after many iterations, F-Boosting can reach a much lower error rate on three out of four test sets. Therefore, in the task of LM for IME where CER is the most important metric, the grid search for feature selection is more desirable.",
"cite_spans": [],
"ref_spans": [
{
"start": 257,
"end": 271,
"text": "Figures 5 to 8",
"ref_id": "FIGREF3"
}
],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "To investigate the impact of using a fixed versus an optimal step size for feature update, we compared FSLR with Boosting. Although both algorithms achieve very similar CER results, the performance of FSLR is much less sensitive to the selected fixed step size. For example, we can select any value from 0.2 to 0.8, and in most settings FSLR achieves the very similar lowest CER after 20,000 iterations, and will stay there for many iterations. In contrast, in Boosting, the optimal value of \u03b5 in Equation (16) varies with the sizes and domains of training data, and has to be tuned carefully. We thus conclude that in our task FSLR is more robust against different training settings and a fixed step size for feature update is more preferred.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dicussion",
"sec_num": "5.3"
},
{
"text": "This paper investigates two approximation lasso methods for LM applied to a realistic task with a very large number of features with sparse feature space. Our results on Japanese text input are promising. BLasso outperforms the boosting algorithm significantly in terms of CER reduction on all experimental settings.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "6"
},
{
"text": "We have shown that this superior performance is a consequence of BLasso's backward step and its fixed step size in both feature selection and feature weight update. Our experimental results in Section 5 show that the use of backward step is vital for model fine-tuning after major features are selected and for coping with strongly correlated features; the fixed step size of BLasso is responsible for the improvement of CER and the robustness of the results. Experiments on other data sets and theoretical analysis are needed to further support our findings in this paper. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "6"
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Language model adaptation with MAP estimation and the perceptron algorithm",
"authors": [
{
"first": "M",
"middle": [],
"last": "Bacchiani",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Roark",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Saraclar",
"suffix": ""
}
],
"year": 2004,
"venue": "HLT-NAACL 2004",
"volume": "",
"issue": "",
"pages": "21--24",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bacchiani, M., Roark, B., and Saraclar, M. 2004. Lan- guage model adaptation with MAP estimation and the perceptron algorithm. In HLT-NAACL 2004. 21-24.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"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": "Collins, Michael and Terry Koo 2005. Discriminative reranking for natural language parsing. Computational Linguistics 31(1): 25-69.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Wavelet shrinkage; asymptopia? (with discussion)",
"authors": [
{
"first": "Richard",
"middle": [
"O"
],
"last": "Duda",
"suffix": ""
},
{
"first": "Peter",
"middle": [
"E"
],
"last": "Hart",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Stork",
"suffix": ""
},
{
"first": "G",
"middle": [
";"
],
"last": "David",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Donoho",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Johnstone",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Kerkyachairan",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Picard",
"suffix": ""
}
],
"year": 1995,
"venue": "J. Royal. Statist. Soc",
"volume": "57",
"issue": "",
"pages": "201--337",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Duda, Richard O, Hart, Peter E. and Stork, David G. 2001. Pattern classification. John Wiley & Sons, Inc. Donoho, D., I. Johnstone, G. Kerkyachairan, and D. Picard. 1995. Wavelet shrinkage; asymptopia? (with discussion), J. Royal. Statist. Soc. 57: 201-337.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Least angle regression",
"authors": [
{
"first": "B",
"middle": [],
"last": "Efron",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Hastie",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Johnstone",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Tibshirani",
"suffix": ""
}
],
"year": 2004,
"venue": "Ann. Statist",
"volume": "32",
"issue": "",
"pages": "407--499",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Efron, B., T. Hastie, I. Johnstone, and R. Tibshirani. 2004. Least angle regression. Ann. Statist. 32, 407-499.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "An efficient boosting algorithm for combining preferences",
"authors": [
{
"first": "Y",
"middle": [],
"last": "Freund",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Iyer",
"suffix": ""
},
{
"first": "R",
"middle": [
"E"
],
"last": "Schapire",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Singer",
"suffix": ""
}
],
"year": 1998,
"venue": "ICML'98",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Freund, Y, R. Iyer, R. E. Schapire, and Y. Singer. 1998. An efficient boosting algorithm for combining pref- erences. In ICML'98.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "The elements of statistical learning",
"authors": [
{
"first": "T",
"middle": [],
"last": "Hastie",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Tibshirani",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Friedman",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hastie, T., R. Tibshirani and J. Friedman. 2001. The elements of statistical learning. Springer-Verlag, New York.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Exploiting headword dependency and predictive clustering for language modeling",
"authors": [
{
"first": "Jianfeng",
"middle": [],
"last": "Gao",
"suffix": ""
},
{
"first": "Hisami",
"middle": [],
"last": "Suzuki",
"suffix": ""
},
{
"first": "Yang",
"middle": [],
"last": "Wen",
"suffix": ""
}
],
"year": 2002,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gao, Jianfeng, Hisami Suzuki and Yang Wen. 2002. Exploiting headword dependency and predictive clustering for language modeling. In EMNLP 2002.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Minimum sample risk methods for language modeling",
"authors": [
{
"first": "",
"middle": [
"J"
],
"last": "Gao",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Yu",
"suffix": ""
},
{
"first": "W",
"middle": [],
"last": "Yuan",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Xu",
"suffix": ""
}
],
"year": 2005,
"venue": "HLT/EMNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gao. J., Yu, H., Yuan, W., and Xu, P. 2005. Minimum sample risk methods for language modeling. In HLT/EMNLP 2005.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "A new approach to variable selection in least squares problems",
"authors": [
{
"first": "M",
"middle": [
"R"
],
"last": "Osborne",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Presnell",
"suffix": ""
},
{
"first": "B",
"middle": [
"A"
],
"last": "Turlach",
"suffix": ""
}
],
"year": 2000,
"venue": "Journal of Numerical Analysis",
"volume": "20",
"issue": "3",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Osborne, M.R. and Presnell, B. and Turlach B.A. 2000a. A new approach to variable selection in least squares problems. Journal of Numerical Analysis, 20(3).",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "On the lasso and its dual",
"authors": [
{
"first": "M",
"middle": [
"R"
],
"last": "Osborne",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Presnell",
"suffix": ""
},
{
"first": "B",
"middle": [
"A"
],
"last": "Turlach",
"suffix": ""
}
],
"year": 2000,
"venue": "Journal of Computational and Graphical Statistics",
"volume": "9",
"issue": "2",
"pages": "319--337",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Osborne, M.R. and Presnell, B. and Turlach B.A. 2000b. On the lasso and its dual. Journal of Computational and Graphical Statistics, 9(2): 319-337.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Corrective language modeling for large vocabulary ASR with the perceptron algorithm",
"authors": [
{
"first": "Brian",
"middle": [],
"last": "Roark",
"suffix": ""
},
{
"first": "Murat",
"middle": [],
"last": "Saraclar",
"suffix": ""
},
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Roark, Brian, Murat Saraclar and Michael Collins. 2004. Corrective language modeling for large vo- cabulary ASR with the perceptron algorithm. In ICASSP 2004.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Improved boosting algorithms using confidence-rated predictions",
"authors": [
{
"first": "Robert",
"middle": [
"E"
],
"last": "Schapire",
"suffix": ""
},
{
"first": "Yoram",
"middle": [],
"last": "Singer",
"suffix": ""
}
],
"year": 1999,
"venue": "Machine Learning",
"volume": "37",
"issue": "",
"pages": "297--336",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Schapire, Robert E. and Yoram Singer. 1999. Improved boosting algorithms using confidence-rated predic- tions. Machine Learning, 37(3): 297-336.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "A comparative study on language model adaptation using new evaluation metrics",
"authors": [
{
"first": "Hisami",
"middle": [],
"last": "Suzuki",
"suffix": ""
},
{
"first": "Jianfeng",
"middle": [],
"last": "Gao",
"suffix": ""
}
],
"year": 2005,
"venue": "HLT/EMNLP",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Suzuki, Hisami and Jianfeng Gao. 2005. A comparative study on language model adaptation using new evaluation metrics. In HLT/EMNLP 2005.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Regression shrinkage and selection via the lasso",
"authors": [
{
"first": "R",
"middle": [],
"last": "Tibshirani",
"suffix": ""
}
],
"year": 1996,
"venue": "J. R. Statist. Soc. B",
"volume": "58",
"issue": "1",
"pages": "267--288",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tibshirani, R. 1996. Regression shrinkage and selection via the lasso. J. R. Statist. Soc. B, 58(1): 267-288.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "An Empirical Study on Language Model Adaptation Using a Metric of Domain Similarity",
"authors": [
{
"first": "W",
"middle": [],
"last": "Yuan",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Gao",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Suzuki",
"suffix": ""
}
],
"year": 2005,
"venue": "IJCNLP 05",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yuan, W., J. Gao and H. Suzuki. 2005. An Empirical Study on Language Model Adaptation Using a Met- ric of Domain Similarity. In IJCNLP 05.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Boosted lasso",
"authors": [
{
"first": "P",
"middle": [],
"last": "Zhao",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Yu",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhao, P. and B. Yu. 2004. Boosted lasso. Tech Report, Statistics Department, U. C. Berkeley.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "1-norm support vector machines. NIPS 16",
"authors": [
{
"first": "J",
"middle": [
"S"
],
"last": "Zhu",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Rosset",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Hastie",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Tibshirani",
"suffix": ""
}
],
"year": 2003,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhu, J. S. Rosset, T. Hastie, and R. Tibshirani. 2003. 1-norm support vector machines. NIPS 16. MIT Press.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"uris": null,
"text": "The boosting algorithm",
"num": null,
"type_str": "figure"
},
"FIGREF1": {
"uris": null,
"text": "L1 curves: models are trained on the E(8K) dataset.",
"num": null,
"type_str": "figure"
},
"FIGREF2": {
"uris": null,
"text": "Test error curves: models are trained on the E(8K) dataset.",
"num": null,
"type_str": "figure"
},
"FIGREF3": {
"uris": null,
"text": "Test error curves: models are trained on the Y(8K) dataset, including headword bigram features.",
"num": null,
"type_str": "figure"
},
"FIGREF4": {
"uris": null,
"text": "Test error curves: models are trained on the T(8K) dataset, including headword bigram features.",
"num": null,
"type_str": "figure"
},
"FIGREF5": {
"uris": null,
"text": "Test error curves: models are trained on the E(8K) dataset, including headword bigram features.",
"num": null,
"type_str": "figure"
},
"FIGREF6": {
"uris": null,
"text": "Test error curves: models are trained on the S(8K) dataset, including headword bigram features.",
"num": null,
"type_str": "figure"
},
"TABREF1": {
"content": "<table><tr><td colspan=\"3\">Domain Entropy vs.Nikkei Baseline</td><td>MAP (over Baseline)</td><td colspan=\"2\">Boosting (over MAP) BLasso (over MAP/Boosting)</td></tr><tr><td>Y (800)</td><td>7.69</td><td>3.70</td><td>3.70 (+0.00)</td><td>3.13 (+15.41)</td><td>3.01 (+18.65/+3.83)</td></tr><tr><td>Y (8K)</td><td>7.69</td><td>3.70</td><td>3.69 (+0.27)</td><td>2.88 (+21.95)</td><td>2.85 (+22.76/+1.04)</td></tr><tr><td>Y (72K)</td><td>7.69</td><td>3.70</td><td>3.69 (+0.27)</td><td>2.78 (+24.66)</td><td>2.73 (+26.02/+1.80)</td></tr><tr><td>T (800)</td><td>7.95</td><td>5.81</td><td>5.81 (+0.00)</td><td>5.69 (+2.07)</td><td>5.63 (+3.10/+1.05)</td></tr><tr><td>T (8K)</td><td>7.95</td><td>5.81</td><td>5.70 (+1.89)</td><td>5.48 (+5.48)</td><td>5.33 (+6.49/+2.74)</td></tr><tr><td>T (72K)</td><td>7.95</td><td>5.81</td><td>5.47 (+5.85)</td><td>5.33 (+2.56)</td><td>5.05 (+7.68/+5.25)</td></tr><tr><td>E (800)</td><td>9.30</td><td>10.24</td><td>9.60 (+6.25)</td><td>9.82 (-2.29)</td><td>9.18 (+4.38/+6.52)</td></tr><tr><td>E (8K)</td><td>9.30</td><td>10.24</td><td>8.64 (+15.63)</td><td>8.54 (+1.16)</td><td>8.04 (+6.94/+5.85)</td></tr><tr><td>E (72K)</td><td>9.30</td><td>10.24</td><td>7.98 (+22.07)</td><td>7.53 (+5.64)</td><td>7.20 (+9.77/+4.38)</td></tr><tr><td>S (800)</td><td>9.40</td><td>12.18</td><td>11.86 (+2.63)</td><td>11.91 (-0.42)</td><td>11.79 (+0.59/+1.01)</td></tr><tr><td>S (8K)</td><td>9.40</td><td>12.18</td><td>11.15 (+8.46)</td><td>11.09 (+0.54)</td><td>10.73 (+3.77/+3.25)</td></tr><tr><td>S (72K)</td><td>9.40</td><td>12.18</td><td>10.76 (+11.66)</td><td>10.25 (+4.74)</td><td>9.64 (+10.41/+5.95)</td></tr></table>",
"text": "CER (%) and CER reduction (%) (Y=Yomiuri; T=TuneUp; E=Encarta; S=-Shincho)",
"html": null,
"type_str": "table",
"num": null
}
}
}
} |