File size: 101,618 Bytes
6fa4bc9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 |
{
"paper_id": "2022",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T07:13:08.785864Z"
},
"title": "On the Impact of Data Augmentation on Downstream Performance in Natural Language Processing",
"authors": [
{
"first": "Itsuki",
"middle": [],
"last": "Okimura",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Tokyo",
"location": {}
},
"email": "okimura@weblab.t.u-tokyo.ac.jp"
},
{
"first": "Machel",
"middle": [],
"last": "Reid",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Tokyo",
"location": {}
},
"email": "machelreid@weblab.t.u-tokyo.ac.jp"
},
{
"first": "Makoto",
"middle": [],
"last": "Kawano",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Tokyo",
"location": {}
},
"email": "kawano@weblab.t.u-tokyo.ac.jp"
},
{
"first": "Yutaka",
"middle": [],
"last": "Matsuo",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "The University of Tokyo",
"location": {}
},
"email": "matsuo@weblab.t.u-tokyo.ac.jp"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "With in the broader scope of machine learning, data augmentation is a common strategy to improve generalization and robustness of machine learning models. While data augmentation has been widely used within computer vision, its use in the NLP has been comparably rather limited. The reason for this is that within NLP, the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner, and effective data augmentation methods are unclear. In this paper, we look to tackle this by evaluating the impact of 12 data augmentation methods on multiple datasets when finetuning pre-trained language models. We find minimal improvements when data sizes are constrained to a few thousand, with performance degradation when data size is increased. We also use various methods to quantify the strength of data augmentations, and find that these values, though weakly correlate with downstream performance, correlate negatively or positively depending on the task. Furthermore, we find a glaring lack of consistently performant data augmentations. This all alludes to the difficulty of data augmentations for NLP tasks and we are inclined to believe that static data augmentations are not broadly applicable given these properties.",
"pdf_parse": {
"paper_id": "2022",
"_pdf_hash": "",
"abstract": [
{
"text": "With in the broader scope of machine learning, data augmentation is a common strategy to improve generalization and robustness of machine learning models. While data augmentation has been widely used within computer vision, its use in the NLP has been comparably rather limited. The reason for this is that within NLP, the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner, and effective data augmentation methods are unclear. In this paper, we look to tackle this by evaluating the impact of 12 data augmentation methods on multiple datasets when finetuning pre-trained language models. We find minimal improvements when data sizes are constrained to a few thousand, with performance degradation when data size is increased. We also use various methods to quantify the strength of data augmentations, and find that these values, though weakly correlate with downstream performance, correlate negatively or positively depending on the task. Furthermore, we find a glaring lack of consistently performant data augmentations. This all alludes to the difficulty of data augmentations for NLP tasks and we are inclined to believe that static data augmentations are not broadly applicable given these properties.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Data augmentation may be useful in situations where the data size is insufficient for the number of parameters in the model, resulting in overtraining (Perez and Wang, 2017) . It has been pointed out that data augmentation does not degrade the expressive power of the model and achieves an improvement in the generalization performance of the model without adjusting the hyperparameters (Hern\u00e1ndez-Garc\u00eda and K\u00f6nig, 2018) . While data augmentation is standard in the field of computer vision, it is not fully used in natural language processing. Two factors can be cited for this. The first reason is that there has been insufficient unified validation of data augmentation methods for a wide range of datasets and data sizes. Another reason is that it is still unclear what kind of data augmentation is effective for learning. In natural language processing, it is difficult to judge whether a data augmentation method is good or bad without relying on experiments, and it is necessary to search for effective data augmentations by trial and error (Feng et al., 2021) . If it is possible to predict whether a data augmentation is effective for learning before training, it would be possible to search for data augmentations more efficiently.",
"cite_spans": [
{
"start": 151,
"end": 173,
"text": "(Perez and Wang, 2017)",
"ref_id": "BIBREF19"
},
{
"start": 387,
"end": 421,
"text": "(Hern\u00e1ndez-Garc\u00eda and K\u00f6nig, 2018)",
"ref_id": "BIBREF12"
},
{
"start": 1049,
"end": 1068,
"text": "(Feng et al., 2021)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper examines the performance impact of data augmentation methods that have been proposed for natural language processing on various datasets. Through this experiment, we will verify whether the data augmentation method can contribute to the improvement of performance on multiple datasets and problem settings. We also use various measures of the strength of a given data augmentation, and investigate its relationship with performance after learning. We find that although data augmentation strength (i.e. how significantly it perturbs the input) is correlated with the change in downstream performance to a given degree, its sign and degree often varies significantly. Based on this, we believe that static data augmentations are not a wise choice for NLP tasks with a reasonable amount of data, and may need to be combined with data-dependent modeling innovations to be broadly applicable to future work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Data Augmentation for NLP Data augmentation has been explored in NLP recently with EDA (Wei and Zou, 2019) , as well as NL-augmenter (Dhole et al., 2021) . Masked language modeling can be considered to be data augmentation (Devlin et al., 2019) , while dictionary-derived augmentation methods have been employed recently for aug-menting multilingual language models with large improvements Reid et al., 2021; Reid and Artetxe, 2022) . However, Longpre et al. (2020) showed that two data augmentation methods in natural language processing had small effects on pre-trained language models. We further expand the scope of this study to examine the performance impact of 12 different data augmentation methods.",
"cite_spans": [
{
"start": 87,
"end": 106,
"text": "(Wei and Zou, 2019)",
"ref_id": "BIBREF27"
},
{
"start": 120,
"end": 153,
"text": "NL-augmenter (Dhole et al., 2021)",
"ref_id": null
},
{
"start": 223,
"end": 244,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF5"
},
{
"start": 390,
"end": 408,
"text": "Reid et al., 2021;",
"ref_id": "BIBREF23"
},
{
"start": 409,
"end": 432,
"text": "Reid and Artetxe, 2022)",
"ref_id": "BIBREF22"
},
{
"start": 444,
"end": 465,
"text": "Longpre et al. (2020)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Evaluating Data Augmentation In the field of computer vision, researchers have been studying what kind of data augmentation contributes to the performance (Taylor and Nitschke, 2018; Perez and Wang, 2017) . And some studies have been done to create metrics on data augmentation and evaluate the relationship with the performance of the model after training. Gontijo-Lopes et al. 2020proposed two indices, affinity and diversity, to quantify how data augmentation improves the generalization of the model, and pointed out that data augmentation methods that are evaluated as having high affinity and diversity will lead to better performance in computer vision. Meanwhile, it is still unclear what characteristics of data augmentation methods are effective in the field of natural language processing.",
"cite_spans": [
{
"start": 155,
"end": 182,
"text": "(Taylor and Nitschke, 2018;",
"ref_id": "BIBREF26"
},
{
"start": 183,
"end": 204,
"text": "Perez and Wang, 2017)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "In this section, we briefly go over metrics we use to evaluate the strength of our data augmentations of a given task as well as strategies for training using data augmentations.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation Metrics and Training Strategies",
"sec_num": "3"
},
{
"text": "In this subsection, we briefly discuss our two training strategies for incorporating data augmentation. Given an i.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Strategy",
"sec_num": "3.1"
},
{
"text": "i.d. dataset D = {(x 1 , y 1 ), (x 2 , y 2 ), . . . , (x N , y N )} containing N examples",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Strategy",
"sec_num": "3.1"
},
{
"text": "where each x i represents an input, and y i represents the assigned label corresponding to x i . Oftentimes, we simply fit a given model on this dataset. However, given a data augmentation function f (x i ) =x i , wherex i represents an augmented input, we can also augment this dataset to improve the diversity of inputs which should hopefully lead to better model generalization and robustness. That is, we now have augmented dataset",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Strategy",
"sec_num": "3.1"
},
{
"text": "D = {(x 1 , y 1 ), . . . , (x N ,\u0177 N )}.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Strategy",
"sec_num": "3.1"
},
{
"text": "We now explain the following finetuning methods: Normal training Finetuning our models on D 1-step training Finetuning our models jointly on augmented datasetD and original dataset D-this method is commonly employed in computer vision. 2-step training To mitigate the distribution shift introduced by the augmentation, but still allowing the model to learn from the augmented dataset, we look at two-step finetuning where we first finetune onD and then finetune on D.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Strategy",
"sec_num": "3.1"
},
{
"text": "We also look to analyse whether there are certain trends among the strength of augmentation methods and their impact on downstream performance. To do this, we measure the strength of augmentation methods using the following metrics:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Strength",
"sec_num": "3.2"
},
{
"text": "Semantic Similarity We use semantic similarity (Cer et al., 2017) as a measure of strength of data augmentation. For example, if a given example is perturbed in a more significant manner, we assume that it's semantic similarity will decrease, therefore indicating a \"stronger\" data augmentation. We use SentenceBERT (Reimers and Gurevych, 2019) to measure the cosine similarity between sentence representation of the original example x i and sentence representation of augmented examplex i .",
"cite_spans": [
{
"start": 47,
"end": 65,
"text": "(Cer et al., 2017)",
"ref_id": "BIBREF0"
},
{
"start": 316,
"end": 344,
"text": "(Reimers and Gurevych, 2019)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Strength",
"sec_num": "3.2"
},
{
"text": "BLEU We use BLEU (Papineni et al., 2002; Post, 2018) as a metric that works on discrete tokens (therefore more sensitive to exact token matches), that is not model dependent as our semantic similarity measure is. That is, a lower BLEU score represents a stronger data augmentation.",
"cite_spans": [
{
"start": 17,
"end": 40,
"text": "(Papineni et al., 2002;",
"ref_id": "BIBREF18"
},
{
"start": 41,
"end": 52,
"text": "Post, 2018)",
"ref_id": "BIBREF20"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Strength",
"sec_num": "3.2"
},
{
"text": "BERTScore We also use text generation metric BERTScore (Zhang* et al., 2020) , which measures cosine-similarity at a token-level, rather than on a sequence-level like our semantic similarity measure.",
"cite_spans": [
{
"start": 55,
"end": 76,
"text": "(Zhang* et al., 2020)",
"ref_id": "BIBREF29"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Strength",
"sec_num": "3.2"
},
{
"text": "In our analyses (Sec. 5), we measure the correlation between these measures and the \u00b1 change in performance.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Strength",
"sec_num": "3.2"
},
{
"text": "In our experiments, we compared the performance of the model when trained with 12 typical data augmentation methods with that of the model trained without data augmentation. Our data augmentations methods are sourced from NL-Augmenter 1 (Dhole et al., 2021) and nlpaug 2 (Ma, 2019) . We provide additional details in Appendix B.",
"cite_spans": [
{
"start": 271,
"end": 281,
"text": "(Ma, 2019)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Augmentation Methods",
"sec_num": "4.1"
},
{
"text": "In experiments, we use three datasets for different language tasks, MRPC (Dolan and Brockett, 2005), SICK (Marelli et al., 2014) , and SST-2 (Socher et al., 2013) . MRPC is a dataset in which the task is to predict whether a sentence-pair is semantically equivalent. SICK is a dataset that contains a task to infer the connotation between a given premise and an explanation. In this experiment, it is a binary classification problem whether the meaning of the explanatory sentence is contained in the meaning of the premise sentence or not. SST-2 is a binary classification problem in which a dataset for sentiment analysis of sentences is created from movie reviews, are classified as positive or negative. For MRPC and SICK, we extended the data to the second sentence in the experiment, and the combination of the first sentence, the extended second sentence pair, and the original label was used as the augmented data set. For SST-2, the combination of the augmented sentence and the original label was used as the augmented data set.",
"cite_spans": [
{
"start": 106,
"end": 128,
"text": "(Marelli et al., 2014)",
"ref_id": "BIBREF15"
},
{
"start": 141,
"end": 162,
"text": "(Socher et al., 2013)",
"ref_id": "BIBREF25"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Datasets",
"sec_num": "4.2"
},
{
"text": "In this experiment, we used the GPT-2 (345M) (Radford et al., 2019) and BERT-large (Devlin et al., 2018) as pre-trained language models. We train models on a single NVIDIA V100 16GB GPU. We measured the performance of training on the original dataset as a baseline, and compared the performance of fine-tuning on the training dataset with the augmented data. We train models until convergence, and perform early stopping where we use a patience of 3 epochs for all models.",
"cite_spans": [
{
"start": 45,
"end": 67,
"text": "(Radford et al., 2019)",
"ref_id": "BIBREF21"
},
{
"start": 83,
"end": 104,
"text": "(Devlin et al., 2018)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "4.3"
},
{
"text": "Performance Changes Due to Data Augmentation Table 1 shows the scores for single-step and 2-step training on the data set with data augmentation (see Appendix D for per-task results). For both training strategies, we also measure the impact of data size, experimenting with various data sizes (10%, 50%, and 100% of the full dataset). When all data was used for training, we found that no data augmentation that improved scores on average for both the language model and the masked language model, except for the 2-step training with BERT with synonym substitution. This indicates that although data-augmentation has the tendency to help at a smaller scale, perhaps mitigating effects of (lack of) data diversity, as the data scale grows we notice that performance degrades where the augmentations most likely add more noise to the dataset.",
"cite_spans": [],
"ref_spans": [
{
"start": 45,
"end": 52,
"text": "Table 1",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5"
},
{
"text": "Relationship between Data Augmentation Intensity and Post-training Performance The correlation coefficients measured by the difference in F1 scores between the data augmentation intensity obtained by the language model and the masked language model and the baseline for each model and learning method are shown in Table2. A positive value indicates that a weaker (i.e. more similar) data augmentation results in better performance. When we use 1-step training, this correlation is generally positive -this indicates that when using naive data combination, then a more similar (i.e. weaker augmentation) is generally more effective. This supports our hypothesis about distribution shift negatively impact augmentation. However, this finding varies significantly when switching to 2-step training depending on model and dataset. Given the relatively strong performance of 2-step training, this indicates that strength of data augmentation can have varying effects when using various training schedules/models.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Results",
"sec_num": "5"
},
{
"text": "When all the original training data was used for training in the three datasets tested in this study, the effect of data augmentation on performance improvement was small, and the performance on the test data deteriorated in many cases. There are two possible reasons for this. The first is that the augmented data may have become noise. It is almost inevitable that data augmentation will result in the augmentation of sentences whose labels cannot be preserved. If some of the augmented sentences are incorrectly labeled, the quality of the dataset will deteriorate to some extent. Therefore, in a setting where a relatively large number of data can be prepared, such as using all the training data, the negative impact of the decrease in data quality is stronger than the positive impact of the increase in the number of data. The second reason is that the knowledge that can be obtained by data augmentation may have already been acquired through prior learning. This is also pointed out by Longpre et al. (2020) . Therefore, for data 1-step GPT2 1-step BERT 2-step GPT-2 2-step BERT 100% 50% 10% 100% 50% 10% 100% 50% 10% 100% 50% 10% baseline Sentence similarity BLEU BERTScore GPT-2 BERT GPT-2 BERT GPT-2 BERT 1-step 2-step 1-step 2-step 1-step 2-step 1-step 2-step 1-step 2-step 1-step 2-step MRPC 0. augmentation in a specific domain, it is possible that data augmentation based on knowledge about the domain, such as substitution based on a list of words that can be substituted in the domain, which cannot be obtained by pre-training with a general corpus, may be effective. On the other hand, when the number of data used for training was limited, we observed some cases where the performance improved even when using a pre-training model. Therefore, in domains where only a few hundred examples are available, performance improvement can be expected by augmenting the existing data.",
"cite_spans": [
{
"start": 995,
"end": 1016,
"text": "Longpre et al. (2020)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [
{
"start": 1051,
"end": 1163,
"text": "1-step BERT 2-step GPT-2 2-step BERT 100% 50% 10% 100% 50% 10% 100% 50% 10% 100% 50% 10% baseline",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "In addition, in 1-step learning, the weaker the data augmentation, the better the performance. However, in 2-step learning, the relationship between the strength of consistent data augmentation and performance depended on the type of data set. This suggests that in 2-step learning, the effective strength of data augmentation may differ depending on the characteristics of the data set. For example, in MRPC, the difference between the data augmentation intensity and the F1 score of the baseline was negatively correlated because even trivial changes are likely to produce data that become noise in learning. In SICK and SST-2, even if some of the content changes, the labels of the sentences are retained as long as the words indicating relevance and emotion remain the same. In this case, the various sentences created by strong data reinforcement in two-stage learning contribute to the learning pro-cess, allowing clean data to be learned in the second half. This may be why the difference between the strength of the data reinforcement and the F1 score from the baseline may have been positively correlated in some cases. Therefore, by comparing the augmentation intensity determined by the proposed index, it may be possible to efficiently search for promising data augmentation methods before actual training. However, more work needs to be done to effectively use these methods in a practical setting.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "In this paper, we observed that most of the data augmentation methods did not improve performance when training on datasets with thousands of examples, but some of them improved performance when training on datasets with hundreds of examples. This suggests that, depending on the task and the data size, data augmentation may be effective even when a pre-trained language model is used for training. We also defined data augmentation intensity, a measure to evaluate whether data augmentation produces sentences that are different from the original sentences, and evaluated the relationship between this measure and the performance after training. As a result, the data augmentation intensity showed different correlations with the change in performance after training depending on the target dataset. For tasks with enough data, this indi-cates the limited applicability and predictability of static data augmentations. In future work, we believe the NLP community should look at modeling or adaptive learning methods (Dery et al., 2022) to account for these differences in data.",
"cite_spans": [
{
"start": 1019,
"end": 1038,
"text": "(Dery et al., 2022)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "7"
},
{
"text": "https://github.com/GEM-benchmark/ NL-Augmenter",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://github.com/makcedward/nlpaug",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "SemEval-2017 task 1: Semantic textual similarity multilingual and crosslingual focused evaluation",
"authors": [
{
"first": "Daniel",
"middle": [],
"last": "Cer",
"suffix": ""
},
{
"first": "Mona",
"middle": [],
"last": "Diab",
"suffix": ""
},
{
"first": "Eneko",
"middle": [],
"last": "Agirre",
"suffix": ""
},
{
"first": "I\u00f1igo",
"middle": [],
"last": "Lopez-Gazpio",
"suffix": ""
},
{
"first": "Lucia",
"middle": [],
"last": "Specia",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)",
"volume": "",
"issue": "",
"pages": "1--14",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Daniel Cer, Mona Diab, Eneko Agirre, I\u00f1igo Lopez- Gazpio, and Lucia Specia. 2017. SemEval-2017 task 1: Semantic textual similarity multilingual and crosslingual focused evaluation. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017), pages 1-14, Vancouver, Canada. Association for Computational Linguistics.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Dict-mlm: Improved multilingual pre-training using bilingual dictionaries",
"authors": [
{
"first": "Aditi",
"middle": [],
"last": "Chaudhary",
"suffix": ""
},
{
"first": "Karthik",
"middle": [],
"last": "Raman",
"suffix": ""
},
{
"first": "Krishna",
"middle": [],
"last": "Srinivasan",
"suffix": ""
},
{
"first": "Jiecao",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Aditi Chaudhary, Karthik Raman, Krishna Srinivasan, and Jiecao Chen. 2020. Dict-mlm: Improved multi- lingual pre-training using bilingual dictionaries.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Unsupervised cross-lingual representation learning at scale",
"authors": [
{
"first": "Alexis",
"middle": [],
"last": "Conneau",
"suffix": ""
},
{
"first": "Kartikay",
"middle": [],
"last": "Khandelwal",
"suffix": ""
},
{
"first": "Naman",
"middle": [],
"last": "Goyal",
"suffix": ""
},
{
"first": "Vishrav",
"middle": [],
"last": "Chaudhary",
"suffix": ""
},
{
"first": "Guillaume",
"middle": [],
"last": "Wenzek",
"suffix": ""
},
{
"first": "Francisco",
"middle": [],
"last": "Guzm\u00e1n",
"suffix": ""
},
{
"first": "Edouard",
"middle": [],
"last": "Grave",
"suffix": ""
},
{
"first": "Myle",
"middle": [],
"last": "Ott",
"suffix": ""
},
{
"first": "Luke",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "8440--8451",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzm\u00e1n, Edouard Grave, Myle Ott, Luke Zettle- moyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Pro- ceedings of the 58th Annual Meeting of the Asso- ciation for Computational Linguistics, pages 8440- 8451, Online. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Ameet Talwalkar, and Graham Neubig. 2022. Should we be pre-training? an argument for end-task aware training as an alternative",
"authors": [
{
"first": "Lucio",
"middle": [
"M"
],
"last": "Dery",
"suffix": ""
},
{
"first": "Paul",
"middle": [],
"last": "Michel",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lucio M. Dery, Paul Michel, Ameet Talwalkar, and Gra- ham Neubig. 2022. Should we be pre-training? an argument for end-task aware training as an alterna- tive.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Bert: Pre-training of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "Jacob",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "Ming-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Kenton",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1810.04805"
]
},
"num": null,
"urls": [],
"raw_text": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understand- ing. arXiv preprint arXiv:1810.04805.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "BERT: Pre-training of deep bidirectional transformers for language understanding",
"authors": [
{
"first": "Jacob",
"middle": [],
"last": "Devlin",
"suffix": ""
},
{
"first": "Ming-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Kenton",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Kristina",
"middle": [],
"last": "Toutanova",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "1",
"issue": "",
"pages": "4171--4186",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language under- standing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Tech- nologies, Volume 1 (Long and Short Papers), pages 4171-4186, Minneapolis, Minnesota. Association for Computational Linguistics.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Timothy Sum Hon Mun",
"authors": [
{
"first": "D",
"middle": [],
"last": "Kaustubh",
"suffix": ""
},
{
"first": "Varun",
"middle": [],
"last": "Dhole",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "Gangal",
"suffix": ""
},
{
"first": "Aadesh",
"middle": [],
"last": "Gehrmann",
"suffix": ""
},
{
"first": "Zhenhao",
"middle": [],
"last": "Gupta",
"suffix": ""
},
{
"first": "Saad",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Abinaya",
"middle": [],
"last": "Mahamood",
"suffix": ""
},
{
"first": "Simon",
"middle": [],
"last": "Mahendiran",
"suffix": ""
},
{
"first": "Ashish",
"middle": [],
"last": "Mille",
"suffix": ""
},
{
"first": "Samson",
"middle": [],
"last": "Srivastava",
"suffix": ""
},
{
"first": "Tongshuang",
"middle": [],
"last": "Tan",
"suffix": ""
},
{
"first": "Jascha",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Jinho",
"middle": [
"D"
],
"last": "Sohl-Dickstein",
"suffix": ""
},
{
"first": "Eduard",
"middle": [],
"last": "Choi",
"suffix": ""
},
{
"first": "Ondrej",
"middle": [],
"last": "Hovy",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "Dusek",
"suffix": ""
},
{
"first": "Sajant",
"middle": [],
"last": "Ruder",
"suffix": ""
},
{
"first": "Nagender",
"middle": [],
"last": "Anand",
"suffix": ""
},
{
"first": "Rabin",
"middle": [],
"last": "Aneja",
"suffix": ""
},
{
"first": "Lisa",
"middle": [],
"last": "Banjade",
"suffix": ""
},
{
"first": "Hanna",
"middle": [],
"last": "Barthe",
"suffix": ""
},
{
"first": "Ian",
"middle": [],
"last": "Behnke",
"suffix": ""
},
{
"first": "Connor",
"middle": [],
"last": "Berlot-Attwell",
"suffix": ""
},
{
"first": "Caroline",
"middle": [],
"last": "Boyle",
"suffix": ""
},
{
"first": "Marco",
"middle": [],
"last": "Brun",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Antonio Sobrevilla",
"suffix": ""
},
{
"first": "Samuel",
"middle": [],
"last": "Cabezudo",
"suffix": ""
},
{
"first": "Emile",
"middle": [],
"last": "Cahyawijaya",
"suffix": ""
},
{
"first": "Wanxiang",
"middle": [],
"last": "Chapuis",
"suffix": ""
},
{
"first": "Mukund",
"middle": [],
"last": "Che",
"suffix": ""
},
{
"first": "Christian",
"middle": [],
"last": "Choudhary",
"suffix": ""
},
{
"first": "Pierre",
"middle": [],
"last": "Clauss",
"suffix": ""
},
{
"first": "Filip",
"middle": [],
"last": "Colombo",
"suffix": ""
},
{
"first": "Gautier",
"middle": [],
"last": "Cornell",
"suffix": ""
},
{
"first": "Mayukh",
"middle": [],
"last": "Dagan",
"suffix": ""
},
{
"first": "Tanay",
"middle": [],
"last": "Das",
"suffix": ""
},
{
"first": "Thomas",
"middle": [],
"last": "Dixit",
"suffix": ""
},
{
"first": "Paul-Alexis",
"middle": [],
"last": "Dopierre",
"suffix": ""
},
{
"first": "Suchitra",
"middle": [],
"last": "Dray",
"suffix": ""
},
{
"first": "Tatiana",
"middle": [],
"last": "Dubey",
"suffix": ""
},
{
"first": "Marco",
"middle": [
"Di"
],
"last": "Ekeinhor",
"suffix": ""
},
{
"first": "Rishabh",
"middle": [],
"last": "Giovanni",
"suffix": ""
},
{
"first": "Rishabh",
"middle": [],
"last": "Gupta",
"suffix": ""
},
{
"first": "Louanes",
"middle": [],
"last": "Gupta",
"suffix": ""
},
{
"first": "Sang",
"middle": [],
"last": "Hamla",
"suffix": ""
},
{
"first": "Fabrice",
"middle": [],
"last": "Han",
"suffix": ""
},
{
"first": "Antoine",
"middle": [],
"last": "Harel-Canada",
"suffix": ""
},
{
"first": "Ishan",
"middle": [],
"last": "Honore",
"suffix": ""
},
{
"first": "Przemyslaw",
"middle": [
"K"
],
"last": "Jindal",
"suffix": ""
},
{
"first": "Denis",
"middle": [],
"last": "Joniak",
"suffix": ""
},
{
"first": "Venelin",
"middle": [],
"last": "Kleyko",
"suffix": ""
},
{
"first": "Kalpesh",
"middle": [],
"last": "Kovatchev",
"suffix": ""
},
{
"first": "Ashutosh",
"middle": [],
"last": "Krishna",
"suffix": ""
},
{
"first": "Stefan",
"middle": [],
"last": "Kumar",
"suffix": ""
},
{
"first": "Seungjae",
"middle": [
"Ryan"
],
"last": "Langer",
"suffix": ""
},
{
"first": "Corey",
"middle": [
"James"
],
"last": "Lee",
"suffix": ""
},
{
"first": "Hualou",
"middle": [],
"last": "Levinson",
"suffix": ""
},
{
"first": "Kaizhao",
"middle": [],
"last": "Liang",
"suffix": ""
},
{
"first": "Zhexiong",
"middle": [],
"last": "Liang",
"suffix": ""
},
{
"first": "Andrey",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Vukosi",
"middle": [],
"last": "Lukyanenko",
"suffix": ""
},
{
"first": "Gerard",
"middle": [],
"last": "Marivate",
"suffix": ""
},
{
"first": "Simon",
"middle": [],
"last": "De Melo",
"suffix": ""
},
{
"first": "Maxime",
"middle": [],
"last": "Meoni",
"suffix": ""
},
{
"first": "Afnan",
"middle": [],
"last": "Meyer",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mir",
"suffix": ""
}
],
"year": null,
"venue": "Jamie Simon",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kaustubh D. Dhole, Varun Gangal, Sebastian Gehrmann, Aadesh Gupta, Zhenhao Li, Saad Ma- hamood, Abinaya Mahendiran, Simon Mille, Ashish Srivastava, Samson Tan, Tongshuang Wu, Jascha Sohl-Dickstein, Jinho D. Choi, Eduard Hovy, On- drej Dusek, Sebastian Ruder, Sajant Anand, Na- gender Aneja, Rabin Banjade, Lisa Barthe, Hanna Behnke, Ian Berlot-Attwell, Connor Boyle, Car- oline Brun, Marco Antonio Sobrevilla Cabezudo, Samuel Cahyawijaya, Emile Chapuis, Wanxiang Che, Mukund Choudhary, Christian Clauss, Pierre Colombo, Filip Cornell, Gautier Dagan, Mayukh Das, Tanay Dixit, Thomas Dopierre, Paul-Alexis Dray, Su- chitra Dubey, Tatiana Ekeinhor, Marco Di Giovanni, Rishabh Gupta, Rishabh Gupta, Louanes Hamla, Sang Han, Fabrice Harel-Canada, Antoine Honore, Ishan Jindal, Przemyslaw K. Joniak, Denis Kleyko, Venelin Kovatchev, Kalpesh Krishna, Ashutosh Ku- mar, Stefan Langer, Seungjae Ryan Lee, Corey James Levinson, Hualou Liang, Kaizhao Liang, Zhexiong Liu, Andrey Lukyanenko, Vukosi Marivate, Gerard de Melo, Simon Meoni, Maxime Meyer, Afnan Mir, Nafise Sadat Moosavi, Niklas Muennighoff, Timo- thy Sum Hon Mun, Kenton Murray, Marcin Namysl, Maria Obedkova, Priti Oli, Nivranshu Pasricha, Jan Pfister, Richard Plant, Vinay Prabhu, Vasile Pais, Libo Qin, Shahab Raji, Pawan Kumar Rajpoot, Vikas Raunak, Roy Rinberg, Nicolas Roberts, Juan Diego Rodriguez, Claude Roux, Vasconcellos P. H. S., Ananya B. Sai, Robin M. Schmidt, Thomas Scialom, Tshephisho Sefara, Saqib N. Shamsi, Xudong Shen, Haoyue Shi, Yiwen Shi, Anna Shvets, Nick Siegel, Damien Sileo, Jamie Simon, Chandan Singh, Ro- man Sitelew, Priyank Soni, Taylor Sorensen, William Soto, Aman Srivastava, KV Aditya Srivatsa, Tony Sun, Mukund Varma T, A Tabassum, Fiona Anting Tan, Ryan Teehan, Mo Tiwari, Marie Tolkiehn, Athena Wang, Zijian Wang, Gloria Wang, Zijie J.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Nl-augmenter: A framework for task-sensitive natural language augmentation",
"authors": [
{
"first": "Fuxuan",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Bryan",
"middle": [],
"last": "Wei",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Wilie",
"suffix": ""
},
{
"first": "Xinyi",
"middle": [],
"last": "Genta Indra Winata",
"suffix": ""
},
{
"first": "Witold",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Tianbao",
"middle": [],
"last": "Wydma\u0144ski",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Xie",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Usama Yaseen",
"suffix": ""
},
{
"first": "Jing",
"middle": [],
"last": "Yee",
"suffix": ""
},
{
"first": "Yue",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Zhang",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wang, Fuxuan Wei, Bryan Wilie, Genta Indra Winata, Xinyi Wu, Witold Wydma\u0144ski, Tianbao Xie, Usama Yaseen, M. Yee, Jing Zhang, and Yue Zhang. 2021. Nl-augmenter: A framework for task-sensitive natu- ral language augmentation.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Automatically constructing a corpus of sentential paraphrases",
"authors": [
{
"first": "B",
"middle": [],
"last": "William",
"suffix": ""
},
{
"first": "Chris",
"middle": [],
"last": "Dolan",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Brockett",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the Third International Workshop on Paraphrasing (IWP2005)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "William B Dolan and Chris Brockett. 2005. Automati- cally constructing a corpus of sentential paraphrases. In Proceedings of the Third International Workshop on Paraphrasing (IWP2005).",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "WordNet: An Electronic Lexical Database",
"authors": [
{
"first": "Christiane",
"middle": [],
"last": "Fellbaum",
"suffix": ""
}
],
"year": 1998,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Christiane Fellbaum. 1998. WordNet: An Electronic Lexical Database. Bradford Books.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Teruko Mitamura, and Eduard Hovy. 2021. A survey of data augmentation approaches for NLP",
"authors": [
{
"first": "Y",
"middle": [],
"last": "Steven",
"suffix": ""
},
{
"first": "Varun",
"middle": [],
"last": "Feng",
"suffix": ""
},
{
"first": "Jason",
"middle": [],
"last": "Gangal",
"suffix": ""
},
{
"first": "Sarath",
"middle": [],
"last": "Wei",
"suffix": ""
},
{
"first": "Soroush",
"middle": [],
"last": "Chandar",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Vosoughi",
"suffix": ""
}
],
"year": null,
"venue": "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
"volume": "",
"issue": "",
"pages": "968--988",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Steven Y. Feng, Varun Gangal, Jason Wei, Sarath Chan- dar, Soroush Vosoughi, Teruko Mitamura, and Ed- uard Hovy. 2021. A survey of data augmentation approaches for NLP. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 968-988, Online. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Affinity and diversity: Quantifying mechanisms of data augmentation",
"authors": [
{
"first": "Raphael",
"middle": [],
"last": "Gontijo-Lopes",
"suffix": ""
},
{
"first": "Sylvia",
"middle": [
"J"
],
"last": "Smullin",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Ekin",
"suffix": ""
},
{
"first": "Ethan",
"middle": [],
"last": "Cubuk",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Dyer",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2002.08973"
]
},
"num": null,
"urls": [],
"raw_text": "Raphael Gontijo-Lopes, Sylvia J Smullin, Ekin D Cubuk, and Ethan Dyer. 2020. Affinity and diver- sity: Quantifying mechanisms of data augmentation. arXiv preprint arXiv:2002.08973.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Data augmentation instead of explicit regularization",
"authors": [
{
"first": "Alex",
"middle": [],
"last": "Hern\u00e1ndez",
"suffix": ""
},
{
"first": "-",
"middle": [],
"last": "Garc\u00eda",
"suffix": ""
},
{
"first": "Peter",
"middle": [],
"last": "K\u00f6nig",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1806.03852"
]
},
"num": null,
"urls": [],
"raw_text": "Alex Hern\u00e1ndez-Garc\u00eda and Peter K\u00f6nig. 2018. Data augmentation instead of explicit regularization. arXiv preprint arXiv:1806.03852.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "How effective is task-agnostic data augmentation for pretrained transformers?",
"authors": [
{
"first": "Shayne",
"middle": [],
"last": "Longpre",
"suffix": ""
},
{
"first": "Yu",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Dubois",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2010.01764"
]
},
"num": null,
"urls": [],
"raw_text": "Shayne Longpre, Yu Wang, and Christopher DuBois. 2020. How effective is task-agnostic data augmen- tation for pretrained transformers? arXiv preprint arXiv:2010.01764.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Nlp augmentation",
"authors": [
{
"first": "Edward",
"middle": [],
"last": "Ma",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Edward Ma. 2019. Nlp augmentation.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "SemEval-2014 task 1: Evaluation of compositional distributional semantic models on full sentences through semantic relatedness and textual entailment",
"authors": [
{
"first": "Marco",
"middle": [],
"last": "Marelli",
"suffix": ""
},
{
"first": "Luisa",
"middle": [],
"last": "Bentivogli",
"suffix": ""
},
{
"first": "Marco",
"middle": [],
"last": "Baroni",
"suffix": ""
},
{
"first": "Raffaella",
"middle": [],
"last": "Bernardi",
"suffix": ""
},
{
"first": "Stefano",
"middle": [],
"last": "Menini",
"suffix": ""
},
{
"first": "Roberto",
"middle": [],
"last": "Zamparelli",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 8th International Workshop on Semantic Evaluation",
"volume": "",
"issue": "",
"pages": "1--8",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Marco Marelli, Luisa Bentivogli, Marco Baroni, Raf- faella Bernardi, Stefano Menini, and Roberto Zam- parelli. 2014. SemEval-2014 task 1: Evaluation of compositional distributional semantic models on full sentences through semantic relatedness and textual entailment. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 1-8, Dublin, Ireland. Association for Compu- tational Linguistics.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Distributed representations of words and phrases and their compositionality",
"authors": [
{
"first": "Tomas",
"middle": [],
"last": "Mikolov",
"suffix": ""
},
{
"first": "Ilya",
"middle": [],
"last": "Sutskever",
"suffix": ""
},
{
"first": "Kai",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Greg",
"middle": [
"S"
],
"last": "Corrado",
"suffix": ""
},
{
"first": "Jeff",
"middle": [],
"last": "Dean",
"suffix": ""
}
],
"year": 2013,
"venue": "Advances in Neural Information Processing Systems",
"volume": "",
"issue": "",
"pages": "3111--3119",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Cor- rado, and Jeff Dean. 2013. Distributed representa- tions of words and phrases and their compositionality. In Advances in Neural Information Processing Sys- tems, pages 3111-3119.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Facebook fair's wmt19 news translation task submission",
"authors": [
{
"first": "Nathan",
"middle": [],
"last": "Ng",
"suffix": ""
},
{
"first": "Kyra",
"middle": [],
"last": "Yee",
"suffix": ""
},
{
"first": "Alexei",
"middle": [],
"last": "Baevski",
"suffix": ""
},
{
"first": "Myle",
"middle": [],
"last": "Ott",
"suffix": ""
},
{
"first": "Michael",
"middle": [],
"last": "Auli",
"suffix": ""
},
{
"first": "Sergey",
"middle": [],
"last": "Edunov",
"suffix": ""
}
],
"year": 2019,
"venue": "Proc. of WMT",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nathan Ng, Kyra Yee, Alexei Baevski, Myle Ott, Michael Auli, and Sergey Edunov. 2019. Facebook fair's wmt19 news translation task submission. In Proc. of WMT.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Bleu: a method for automatic evaluation of machine translation",
"authors": [
{
"first": "Kishore",
"middle": [],
"last": "Papineni",
"suffix": ""
},
{
"first": "Salim",
"middle": [],
"last": "Roukos",
"suffix": ""
},
{
"first": "Todd",
"middle": [],
"last": "Ward",
"suffix": ""
},
{
"first": "Wei-Jing",
"middle": [],
"last": "Zhu",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "311--318",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu. 2002. Bleu: a method for automatic evalu- ation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Compu- tational Linguistics, pages 311-318, Philadelphia, Pennsylvania, USA. Association for Computational Linguistics.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "The effectiveness of data augmentation in image classification using deep learning",
"authors": [
{
"first": "Luis",
"middle": [],
"last": "Perez",
"suffix": ""
},
{
"first": "Jason",
"middle": [],
"last": "Wang",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1712.04621"
]
},
"num": null,
"urls": [],
"raw_text": "Luis Perez and Jason Wang. 2017. The effectiveness of data augmentation in image classification using deep learning. arXiv preprint arXiv:1712.04621.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "A call for clarity in reporting BLEU scores",
"authors": [
{
"first": "Matt",
"middle": [],
"last": "Post",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the Third Conference on Machine Translation: Research Papers",
"volume": "",
"issue": "",
"pages": "186--191",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Matt Post. 2018. A call for clarity in reporting BLEU scores. In Proceedings of the Third Conference on Machine Translation: Research Papers, pages 186- 191, Brussels, Belgium. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Language models are unsupervised multitask learners",
"authors": [
{
"first": "Alec",
"middle": [],
"last": "Radford",
"suffix": ""
},
{
"first": "Jeffrey",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Rewon",
"middle": [],
"last": "Child",
"suffix": ""
},
{
"first": "David",
"middle": [],
"last": "Luan",
"suffix": ""
},
{
"first": "Dario",
"middle": [],
"last": "Amodei",
"suffix": ""
},
{
"first": "Ilya",
"middle": [],
"last": "Sutskever",
"suffix": ""
}
],
"year": 2019,
"venue": "OpenAI blog",
"volume": "1",
"issue": "8",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "PARADISE: Exploiting parallel data for multilingual sequenceto-sequence pretraining",
"authors": [
{
"first": "Machel",
"middle": [],
"last": "Reid",
"suffix": ""
},
{
"first": "Mikel",
"middle": [],
"last": "Artetxe",
"suffix": ""
}
],
"year": 2022,
"venue": "Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Machel Reid and Mikel Artetxe. 2022. PARADISE: Exploiting parallel data for multilingual sequence- to-sequence pretraining. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "AfroMT: Pretraining strategies and reproducible benchmarks for translation of 8 African languages",
"authors": [
{
"first": "Machel",
"middle": [],
"last": "Reid",
"suffix": ""
},
{
"first": "Junjie",
"middle": [],
"last": "Hu",
"suffix": ""
},
{
"first": "Graham",
"middle": [],
"last": "Neubig",
"suffix": ""
},
{
"first": "Yutaka",
"middle": [],
"last": "Matsuo",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "1306--1320",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Machel Reid, Junjie Hu, Graham Neubig, and Yutaka Matsuo. 2021. AfroMT: Pretraining strategies and reproducible benchmarks for translation of 8 African languages. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Process- ing, pages 1306-1320, Online and Punta Cana, Do- minican Republic. Association for Computational Linguistics.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "Sentence-bert: Sentence embeddings using siamese bert-networks",
"authors": [
{
"first": "Nils",
"middle": [],
"last": "Reimers",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Recursive deep models for semantic compositionality over a sentiment treebank",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Socher",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Perelygin",
"suffix": ""
},
{
"first": "Jean",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Jason",
"middle": [],
"last": "Chuang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Christopher",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Manning",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Andrew",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Ng",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Potts",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the 2013 conference on empirical methods in natural language processing",
"volume": "",
"issue": "",
"pages": "1631--1642",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D Manning, Andrew Y Ng, and Christopher Potts. 2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proceedings of the 2013 conference on empiri- cal methods in natural language processing, pages 1631-1642.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Improving deep learning with generic data augmentation",
"authors": [
{
"first": "Luke",
"middle": [],
"last": "Taylor",
"suffix": ""
},
{
"first": "Geoff",
"middle": [],
"last": "Nitschke",
"suffix": ""
}
],
"year": 2018,
"venue": "2018 IEEE Symposium Series on Computational Intelligence (SSCI)",
"volume": "",
"issue": "",
"pages": "1542--1547",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Luke Taylor and Geoff Nitschke. 2018. Improving deep learning with generic data augmentation. In 2018 IEEE Symposium Series on Computational Intelli- gence (SSCI), pages 1542-1547. IEEE.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "EDA: Easy data augmentation techniques for boosting performance on text classification tasks",
"authors": [
{
"first": "Jason",
"middle": [],
"last": "Wei",
"suffix": ""
},
{
"first": "Kai",
"middle": [],
"last": "Zou",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
"volume": "",
"issue": "",
"pages": "6382--6388",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jason Wei and Kai Zou. 2019. EDA: Easy data augmen- tation techniques for boosting performance on text classification tasks. In Proceedings of the 2019 Con- ference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP- IJCNLP), pages 6382-6388, Hong Kong, China. As- sociation for Computational Linguistics.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Transformers: State-of-the-art natural language processing",
"authors": [
{
"first": "Thomas",
"middle": [],
"last": "Wolf",
"suffix": ""
},
{
"first": "Lysandre",
"middle": [],
"last": "Debut",
"suffix": ""
},
{
"first": "Victor",
"middle": [],
"last": "Sanh",
"suffix": ""
},
{
"first": "Julien",
"middle": [],
"last": "Chaumond",
"suffix": ""
},
{
"first": "Clement",
"middle": [],
"last": "Delangue",
"suffix": ""
},
{
"first": "Anthony",
"middle": [],
"last": "Moi",
"suffix": ""
},
{
"first": "Pierric",
"middle": [],
"last": "Cistac",
"suffix": ""
},
{
"first": "Tim",
"middle": [],
"last": "Rault",
"suffix": ""
},
{
"first": "Remi",
"middle": [],
"last": "Louf",
"suffix": ""
},
{
"first": "Morgan",
"middle": [],
"last": "Funtowicz",
"suffix": ""
},
{
"first": "Joe",
"middle": [],
"last": "Davison",
"suffix": ""
},
{
"first": "Sam",
"middle": [],
"last": "Shleifer",
"suffix": ""
},
{
"first": "Clara",
"middle": [],
"last": "Patrick Von Platen",
"suffix": ""
},
{
"first": "Yacine",
"middle": [],
"last": "Ma",
"suffix": ""
},
{
"first": "Julien",
"middle": [],
"last": "Jernite",
"suffix": ""
},
{
"first": "Canwen",
"middle": [],
"last": "Plu",
"suffix": ""
},
{
"first": "Teven",
"middle": [
"Le"
],
"last": "Xu",
"suffix": ""
},
{
"first": "Sylvain",
"middle": [],
"last": "Scao",
"suffix": ""
},
{
"first": "Mariama",
"middle": [],
"last": "Gugger",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Drame",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
"volume": "",
"issue": "",
"pages": "38--45",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier- ric Cistac, Tim Rault, Remi Louf, Morgan Funtow- icz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoest, and Alexander Rush. 2020. Trans- formers: State-of-the-art natural language processing. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 38-45, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "Bertscore: Evaluating text generation with bert",
"authors": [
{
"first": "Tianyi",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "*",
"middle": [],
"last": "",
"suffix": ""
},
{
"first": "Varsha",
"middle": [],
"last": "Kishore",
"suffix": ""
},
{
"first": "*",
"middle": [],
"last": "",
"suffix": ""
},
{
"first": "Felix",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "*",
"middle": [],
"last": "",
"suffix": ""
},
{
"first": "Kilian",
"middle": [
"Q"
],
"last": "Weinberger",
"suffix": ""
},
{
"first": "Yoav",
"middle": [],
"last": "Artzi",
"suffix": ""
}
],
"year": 2020,
"venue": "International Conference on Learning Representations",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tianyi Zhang*, Varsha Kishore*, Felix Wu*, Kilian Q. Weinberger, and Yoav Artzi. 2020. Bertscore: Eval- uating text generation with bert. In International Conference on Learning Representations.",
"links": null
}
},
"ref_entries": {
"FIGREF1": {
"text": "2478 0.5813 0.5540 0.2011 0.3782 0.5150 0.6574 0.2712 0.2406 0.6119 0.4879 0.2064 SICK 0.5138 -0.1941 0.4790 -0.5216 0.2424 0.4192 0.0392 -0.5645 0.1085 0.1314 0.0483 -0.2592 SST-2 0.3251 0.3216 0.5897 -0.4152 0.2226 0.4876 0.2015 -0.2712 0.1686 0.3699 0.4524 -0.4342",
"uris": null,
"num": null,
"type_str": "figure"
},
"TABREF0": {
"text": "Table ofaverage F1 scores in 1-step and 2-step training for each percentage of data used for training when data augmentation is used for MRPC, SICK and SST-2.",
"num": null,
"html": null,
"content": "<table/>",
"type_str": "table"
},
"TABREF1": {
"text": "Correlation coefficient between data augmentation strength and difference in F1 score from baseline.",
"num": null,
"html": null,
"content": "<table/>",
"type_str": "table"
}
}
}
} |