File size: 224,579 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 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 |
{
"paper_id": "2021",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T03:14:09.853247Z"
},
"title": "Shaking Syntactic Trees on the Sesame Street: Multilingual Probing with Controllable Perturbations",
"authors": [
{
"first": "Ekaterina",
"middle": [],
"last": "Taktasheva",
"suffix": "",
"affiliation": {},
"email": "etaktasheva@hse.ru"
},
{
"first": "Vladislav",
"middle": [],
"last": "Mikhailov",
"suffix": "",
"affiliation": {},
"email": ""
},
{
"first": "Ekaterina",
"middle": [],
"last": "Artemova",
"suffix": "",
"affiliation": {},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Recent research has adopted a new experimental field centered around the concept of text perturbations which has revealed that shuffled word order has little to no impact on the downstream performance of Transformer-based language models across many NLP tasks. These findings contradict the common understanding of how the models encode hierarchical and structural information and even question if the word order is modeled with position embeddings. To this end, this paper proposes nine probing datasets organized by the type of controllable text perturbation for three Indo-European languages with a varying degree of word order flexibility: English, Swedish and Russian. Based on the probing analysis of the M-BERT and M-BART models, we report that the syntactic sensitivity depends on the language and model pre-training objectives. We also find that the sensitivity grows across layers together with the increase of the perturbation granularity. Last but not least, we show that the models barely use the positional information to induce syntactic trees from their intermediate self-attention and contextualized representations.",
"pdf_parse": {
"paper_id": "2021",
"_pdf_hash": "",
"abstract": [
{
"text": "Recent research has adopted a new experimental field centered around the concept of text perturbations which has revealed that shuffled word order has little to no impact on the downstream performance of Transformer-based language models across many NLP tasks. These findings contradict the common understanding of how the models encode hierarchical and structural information and even question if the word order is modeled with position embeddings. To this end, this paper proposes nine probing datasets organized by the type of controllable text perturbation for three Indo-European languages with a varying degree of word order flexibility: English, Swedish and Russian. Based on the probing analysis of the M-BERT and M-BART models, we report that the syntactic sensitivity depends on the language and model pre-training objectives. We also find that the sensitivity grows across layers together with the increase of the perturbation granularity. Last but not least, we show that the models barely use the positional information to induce syntactic trees from their intermediate self-attention and contextualized representations.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "An extensive body of works is devoted to analyzing syntactic knowledge of Transformer language models (LMs) (Vaswani et al., 2017; Clark et al., 2019; Goldberg, 2019; Belinkov and Glass, 2019) . BERTbased LMs (Devlin et al., 2019) have demonstrated their abilities to encode various linguistic and hierarchical properties (Lin et al., 2019; Jawahar et al., 2019; Jo and Myaeng, 2020) which have a positive effect on the downstream performance (Liu et al., 2019a; Miaschi et al., 2020) and serve as an inspiration for syntax-oriented architecture improvements Bai et al., 2021; Ahmad et al., 2021; Sachan et al., 2021) . Besides, a variety of pre-training objectives has been introduced (Liu et al., 2020a) , with some of them modeling reconstruction of the perturbed word order (Lewis et al., 2020; Tao et al., 2021; Panda et al., 2021) .",
"cite_spans": [
{
"start": 108,
"end": 130,
"text": "(Vaswani et al., 2017;",
"ref_id": "BIBREF69"
},
{
"start": 131,
"end": 150,
"text": "Clark et al., 2019;",
"ref_id": "BIBREF10"
},
{
"start": 151,
"end": 166,
"text": "Goldberg, 2019;",
"ref_id": "BIBREF22"
},
{
"start": 167,
"end": 192,
"text": "Belinkov and Glass, 2019)",
"ref_id": "BIBREF7"
},
{
"start": 209,
"end": 230,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF13"
},
{
"start": 322,
"end": 340,
"text": "(Lin et al., 2019;",
"ref_id": "BIBREF37"
},
{
"start": 341,
"end": 362,
"text": "Jawahar et al., 2019;",
"ref_id": "BIBREF29"
},
{
"start": 363,
"end": 383,
"text": "Jo and Myaeng, 2020)",
"ref_id": "BIBREF30"
},
{
"start": 443,
"end": 462,
"text": "(Liu et al., 2019a;",
"ref_id": "BIBREF38"
},
{
"start": 463,
"end": 484,
"text": "Miaschi et al., 2020)",
"ref_id": "BIBREF45"
},
{
"start": 559,
"end": 576,
"text": "Bai et al., 2021;",
"ref_id": "BIBREF4"
},
{
"start": 577,
"end": 596,
"text": "Ahmad et al., 2021;",
"ref_id": "BIBREF0"
},
{
"start": 597,
"end": 617,
"text": "Sachan et al., 2021)",
"ref_id": "BIBREF57"
},
{
"start": 686,
"end": 705,
"text": "(Liu et al., 2020a)",
"ref_id": null
},
{
"start": 778,
"end": 798,
"text": "(Lewis et al., 2020;",
"ref_id": "BIBREF36"
},
{
"start": 799,
"end": 816,
"text": "Tao et al., 2021;",
"ref_id": "BIBREF67"
},
{
"start": 817,
"end": 836,
"text": "Panda et al., 2021)",
"ref_id": "BIBREF48"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Recent research has adopted a new experimental direction aimed at exploring the syntactic knowledge of LMs and their sensitivity to word order employing text perturbations (Futrell et al., 2018 (Futrell et al., , 2019 Ettinger, 2020) . Some studies show that shuffling word order causes significant performance drops on a wide range of QA tasks Sugawara et al., 2020) . However, a number of works demonstrates that such permutation has little to no impact during the pre-training and finetuning stages (Pham et al., 2020; Sinha et al., 2020 Sinha et al., , 2021 O'Connor and Andreas, 2021; Hessel and Schofield, 2021; Gupta et al., 2021) . The latter contradict the common understanding on how the hierarchical and structural information is encoded in LMs (Rogers et al., 2020) , and even may question if the word order is modeled with the position embeddings Dufter et al., 2021) .",
"cite_spans": [
{
"start": 172,
"end": 193,
"text": "(Futrell et al., 2018",
"ref_id": "BIBREF18"
},
{
"start": 194,
"end": 217,
"text": "(Futrell et al., , 2019",
"ref_id": "BIBREF19"
},
{
"start": 218,
"end": 233,
"text": "Ettinger, 2020)",
"ref_id": "BIBREF17"
},
{
"start": 345,
"end": 367,
"text": "Sugawara et al., 2020)",
"ref_id": "BIBREF66"
},
{
"start": 502,
"end": 521,
"text": "(Pham et al., 2020;",
"ref_id": "BIBREF50"
},
{
"start": 522,
"end": 540,
"text": "Sinha et al., 2020",
"ref_id": "BIBREF65"
},
{
"start": 541,
"end": 561,
"text": "Sinha et al., , 2021",
"ref_id": "BIBREF64"
},
{
"start": 562,
"end": 589,
"text": "O'Connor and Andreas, 2021;",
"ref_id": "BIBREF47"
},
{
"start": 590,
"end": 617,
"text": "Hessel and Schofield, 2021;",
"ref_id": "BIBREF24"
},
{
"start": 618,
"end": 637,
"text": "Gupta et al., 2021)",
"ref_id": "BIBREF23"
},
{
"start": 756,
"end": 777,
"text": "(Rogers et al., 2020)",
"ref_id": "BIBREF55"
},
{
"start": 860,
"end": 880,
"text": "Dufter et al., 2021)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This has stimulated a targeted probing of the LMs internal representations generated from original texts and their permuted counterparts (Sinha et al., 2021; Hessel and Schofield, 2021) . A new type of controllable probes has been proposed, designed to test the LMs sensitivity to granular character-and sub-word level manipulations (Clouatre et al., 2021) , as well as structured syntactic perturbations (Alleman et al., 2021) . Despite the emerging interest in the field, little is investigated for languages other than English, specifically those with flexible word order.",
"cite_spans": [
{
"start": 137,
"end": 157,
"text": "(Sinha et al., 2021;",
"ref_id": "BIBREF64"
},
{
"start": 158,
"end": 185,
"text": "Hessel and Schofield, 2021)",
"ref_id": "BIBREF24"
},
{
"start": 333,
"end": 356,
"text": "(Clouatre et al., 2021)",
"ref_id": null
},
{
"start": 405,
"end": 427,
"text": "(Alleman et al., 2021)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper extends the ongoing research on the syntactic sensitivity to three Indo-European languages with a varying degree of word order flexibility: English, Swedish, and Russian. The contributions of this work are summarized as follows. First, we propose nine probing datasets in the languages mentioned above, organized by the type of controllable syntactic perturbation: N-gram perturbation (NgramShift), shuffling parts of the syntactic clauses (ClauseShift) and randomizing word order (RandomShift) . Despite that randomizing word order has been studied from many perspectives (see Section 2), NgramShift differs from similar approaches (Conneau et al., 2018; Ravishankar et al., 2019; Eger et al., 2020; Alleman et al., 2021) in that the N-grams correspond to only syntactic phrases (e.g. prepositional or numerical phrases) rather than random word spans. ClauseShift is a previously unexplored type of syntactic perturbation adopted from the syntactic tree augmentation method (\u015eahin and Steedman, 2018) . Second, we apply a combination of parameter-free interpretation methods to test the sensitivity of two multilingual Transformer LMs: M-BERT (Devlin et al., 2019) , and M-BART (Liu et al., 2020b) . We hypothesize that M-BART is more robust to the perturbations as opposed to M-BERT since it is learned to restore the shuffled input during pre-training. We evaluate the discrepancy in the syntactic trees induced by the models from perturbed sentences against the original ones, along with the ability to distinguish between them by judging their linguistic acceptability (Lau et al., 2020) . Finally, we analyze the relationship between the models' probe performance and position embeddings (PEs). To the best of our knowledge, it is one of the first attempts to introspect PEs regarding structural probing, particularly in the light of syntactic perturbations. The code and datasets are publicly available 1 .",
"cite_spans": [
{
"start": 492,
"end": 505,
"text": "(RandomShift)",
"ref_id": null
},
{
"start": 644,
"end": 666,
"text": "(Conneau et al., 2018;",
"ref_id": "BIBREF12"
},
{
"start": 667,
"end": 692,
"text": "Ravishankar et al., 2019;",
"ref_id": "BIBREF54"
},
{
"start": 693,
"end": 711,
"text": "Eger et al., 2020;",
"ref_id": "BIBREF16"
},
{
"start": 712,
"end": 733,
"text": "Alleman et al., 2021)",
"ref_id": "BIBREF3"
},
{
"start": 986,
"end": 1012,
"text": "(\u015eahin and Steedman, 2018)",
"ref_id": null
},
{
"start": 1155,
"end": 1176,
"text": "(Devlin et al., 2019)",
"ref_id": "BIBREF13"
},
{
"start": 1190,
"end": 1209,
"text": "(Liu et al., 2020b)",
"ref_id": "BIBREF40"
},
{
"start": 1585,
"end": 1603,
"text": "(Lau et al., 2020)",
"ref_id": "BIBREF34"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Syntax Probing Most of the previous studies on the syntactic knowledge of LMs are centered around the concept of probing tasks, where a simple classifier is trained to predict a particular linguistic property based on the model internal representations (Conneau et al., 2018) . The scope of the properties ranges from dependency relations (Tenney et al., 2018) to the depth of a syntax tree, and top constituents (Conneau et al., 2018) . A variety of probing datasets and benchmarks have been developed. To name a few, Liu et al. (2019a) create a probing suite focused on fine-grained linguistic phenomena, including hierarchical knowledge. SyntaxGym (Gauthier et al., 2020) and LIN-SPECTOR (\u015eahin et al., 2020) allow for targeted evaluation of the LMs linguistic knowledge in a standardized and reproducible environment.",
"cite_spans": [
{
"start": 253,
"end": 275,
"text": "(Conneau et al., 2018)",
"ref_id": "BIBREF12"
},
{
"start": 339,
"end": 360,
"text": "(Tenney et al., 2018)",
"ref_id": "BIBREF68"
},
{
"start": 413,
"end": 435,
"text": "(Conneau et al., 2018)",
"ref_id": "BIBREF12"
},
{
"start": 519,
"end": 537,
"text": "Liu et al. (2019a)",
"ref_id": "BIBREF38"
},
{
"start": 651,
"end": 674,
"text": "(Gauthier et al., 2020)",
"ref_id": "BIBREF20"
},
{
"start": 679,
"end": 711,
"text": "LIN-SPECTOR (\u015eahin et al., 2020)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "These studies have proved that LMs are capable of encoding linguistic and hierarchical information (Belinkov and Glass, 2019; Rogers et al., 2020) . However, the probing paradigm has been lately criticized for relying on supervised probes, which can learn linguistic properties given the supervision, and make it challenging to interpret the results because of the additional set of parameters (Hewitt and Liang, 2019; Belinkov, 2021) . Towards that end, Hewitt and Manning (2019) introduce a structural probe to explore a linear transformation of the embedding space, which best approximates the distance between words and depth of the parse tree. The method has proved to infer the hierarchical structure without any linguistic annotation (Kim et al., 2020) . Maudslay and Cotterell 2021propose a Jabberwocky probing suite of semantically nonsensical but syntactically well-formed sentences. The results demonstrate that the BERTbased LMs do not isolate semantics from syntax, which motivates further development of the probing field.",
"cite_spans": [
{
"start": 99,
"end": 125,
"text": "(Belinkov and Glass, 2019;",
"ref_id": "BIBREF7"
},
{
"start": 126,
"end": 146,
"text": "Rogers et al., 2020)",
"ref_id": "BIBREF55"
},
{
"start": 394,
"end": 418,
"text": "(Hewitt and Liang, 2019;",
"ref_id": "BIBREF25"
},
{
"start": 419,
"end": 434,
"text": "Belinkov, 2021)",
"ref_id": "BIBREF6"
},
{
"start": 455,
"end": 480,
"text": "Hewitt and Manning (2019)",
"ref_id": "BIBREF26"
},
{
"start": 741,
"end": 759,
"text": "(Kim et al., 2020)",
"ref_id": "BIBREF32"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Acceptability Judgements Another line of works relies on the concept of acceptability judgments. The CoLA benchmark and its counterpart for Swedish (Volodina et al., 2021) test LMs ability to identify various linguistic violations. Although Transformer LMs have outperformed the CoLA human solvers on the GLUE leaderboard , a granular linguistic analysis (Warstadt and Bowman, 2019) shows that the models struggle with long-distance syntactic phenomena as opposed to more local ones. Similar in spirit, BLiMP (Warstadt et al., 2020) , and CLiMP (Xiang et al., 2021) allow to evaluate the LMs with respect to the acceptability contrasts, framing the task as ranking sentences in minimal pairs.",
"cite_spans": [
{
"start": 148,
"end": 171,
"text": "(Volodina et al., 2021)",
"ref_id": "BIBREF71"
},
{
"start": 355,
"end": 382,
"text": "(Warstadt and Bowman, 2019)",
"ref_id": "BIBREF76"
},
{
"start": 509,
"end": 532,
"text": "(Warstadt et al., 2020)",
"ref_id": "BIBREF77"
},
{
"start": 545,
"end": 565,
"text": "(Xiang et al., 2021)",
"ref_id": "BIBREF81"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Text Perturbations Recent research has adopted a scope of novel approaches to investigating the LMs sensitivity to syntax corruption and input data manipulations. Starting from studies on randomized word order in LSTMs (Hill et al., 2016; Khandelwal et al., 2018; Sankar et al., 2019; Nie et al., 2019 ), text perturbations have emerged as an audacious experimental direction under the \"pre-train & fine-tune\" paradigm along with the interpreta-tion methods of modern LMs. ; Sugawara et al. (2020) show that N-gram permutations and shuffled word order in the fine-tuning data cause BERT's performance drops up to 22% on a wide range of QA tasks. In contrast, several works report that models fine-tuned on such perturbed data still produce high confidence predictions and perform close to their counterparts on many tasks, including the GLUE benchmark (Ahmad et al., 2019; Sinha et al., 2020; Liu et al., 2021; Hessel and Schofield, 2021; Gupta et al., 2021) . Similar results are demonstrated by the RoBERTa model (Liu et al., 2019b) when the word order perturbations are incorporated into the pretraining objective (Panda et al., 2021) or tested as a part of full pre-training on the perturbed corpora (Sinha et al., 2021) . Sinha et al. (2021) find that the randomized RoBERTa models are similar to their naturally pre-trained peer according to parametric probes but perform worse according to the non-parametric ones.",
"cite_spans": [
{
"start": 219,
"end": 238,
"text": "(Hill et al., 2016;",
"ref_id": "BIBREF27"
},
{
"start": 239,
"end": 263,
"text": "Khandelwal et al., 2018;",
"ref_id": "BIBREF31"
},
{
"start": 264,
"end": 284,
"text": "Sankar et al., 2019;",
"ref_id": "BIBREF62"
},
{
"start": 285,
"end": 301,
"text": "Nie et al., 2019",
"ref_id": "BIBREF46"
},
{
"start": 475,
"end": 497,
"text": "Sugawara et al. (2020)",
"ref_id": "BIBREF66"
},
{
"start": 852,
"end": 872,
"text": "(Ahmad et al., 2019;",
"ref_id": "BIBREF1"
},
{
"start": 873,
"end": 892,
"text": "Sinha et al., 2020;",
"ref_id": "BIBREF65"
},
{
"start": 893,
"end": 910,
"text": "Liu et al., 2021;",
"ref_id": "BIBREF42"
},
{
"start": 911,
"end": 938,
"text": "Hessel and Schofield, 2021;",
"ref_id": "BIBREF24"
},
{
"start": 939,
"end": 958,
"text": "Gupta et al., 2021)",
"ref_id": "BIBREF23"
},
{
"start": 1015,
"end": 1034,
"text": "(Liu et al., 2019b)",
"ref_id": "BIBREF41"
},
{
"start": 1117,
"end": 1137,
"text": "(Panda et al., 2021)",
"ref_id": "BIBREF48"
},
{
"start": 1204,
"end": 1224,
"text": "(Sinha et al., 2021)",
"ref_id": "BIBREF64"
},
{
"start": 1227,
"end": 1246,
"text": "Sinha et al. (2021)",
"ref_id": "BIBREF64"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Recognizing the need to further explore the LMs sensitivity to word order, Clouatre et al. (2021) and Alleman et al. (2021) conduct the interpretation analysis of LMs by means of controllable text perturbations. Clouatre et al. (2021) propose two metrics that score local and global structure of sentences perturbed at the granularity of characters and sub-words. The metrics allow identifying that both conventional and Transformer LMs rely on the local order of tokens more than the global one. Alleman et al. (2021) find that BERT builds syntactic complexity towards the output layer and demonstrates a growing sensitivity to the hierarchical phrase structure across layers. In line with these studies, we analyze the syntactic sensitivity of Transformer-based LMs, extending the experimental setup to the multilingual setting.",
"cite_spans": [
{
"start": 75,
"end": 97,
"text": "Clouatre et al. (2021)",
"ref_id": null
},
{
"start": 102,
"end": 123,
"text": "Alleman et al. (2021)",
"ref_id": "BIBREF3"
},
{
"start": 212,
"end": 234,
"text": "Clouatre et al. (2021)",
"ref_id": null
},
{
"start": 497,
"end": 518,
"text": "Alleman et al. (2021)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "This work proposes three types of controllable syntactic perturbations varying in the extent of sentence corruption. We construct nine probing tasks 2 for three Indo-European languages 3 : English (West Germanic, analytic), Swedish (North Germanic, analytic), and Russian (Balto-Slavic, fusional). Based on the dominant constituent order, all three languages are classified as the SVO (Subject-Verb-Object) languages. Nevertheless, there are some differences between them regarding word order flexibility. Russian is known to exhibit free word order as all of the possible constituent reorderings are acceptable: SOV, OSV, SVO, OVS, VSO, VOS (Bailyn, 2012). English allows for only two of them, namely SVO and OSV (Prince, 1988) . Swedish belongs to the verb-second languages, which poses different restrictions on the possible constituent reorderings (B\u00f6rjars et al., 2003) . Each dataset 4 consists of 10k pairs of the corresponding perturbed sentence and its original.",
"cite_spans": [
{
"start": 714,
"end": 728,
"text": "(Prince, 1988)",
"ref_id": "BIBREF51"
},
{
"start": 852,
"end": 874,
"text": "(B\u00f6rjars et al., 2003)",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Controllable Perturbations",
"sec_num": "3"
},
{
"text": "NgramShift tests the LM sensitivity to local perturbations taking into account the syntactic structure. We used a set of carefully designed morphosyntactic patterns to perturb N-grams that correspond to only syntactic phrases such as numeral phrases, determiner phrases, compound noun phrases, prepositional phrases, etc. Towards this, we applied TF-IDF weighting from scikit-learn library (Pedregosa et al., 2011) to build a ranked N-gram feature matrix from the corpora and further used it for the Ngram inversion. We used the N-gram range \u2208 [2; 4] for each language. Note that the number of words that change their absolute positions is similar for different values of N . Figure 1 illustrates the shift of the head in the prepositional phrase \"to school\" for the sentence \"He did not go to school yesterday\". ClauseShift probes the LM sensitivity to distant perturbations at the level of syntactic clauses. We use the syntactic tree augmentation method (\u015eahin and Steedman, 2019) to rotate sub-trees around the root of the dependency tree of each sentence to form a new synthetic sentence. We then apply a set of manually curated language-specific heuristics to filter out sentences uncorrupted by the rotation procedure. Figure 2 outlines an example of the clause rotation perturbation for the sentence \"He manages to tell her that she has been resurrected\".",
"cite_spans": [
{
"start": 390,
"end": 414,
"text": "(Pedregosa et al., 2011)",
"ref_id": "BIBREF49"
}
],
"ref_spans": [
{
"start": 676,
"end": 684,
"text": "Figure 1",
"ref_id": "FIGREF1"
},
{
"start": 1226,
"end": 1234,
"text": "Figure 2",
"ref_id": "FIGREF11"
}
],
"eq_spans": [],
"section": "Controllable Perturbations",
"sec_num": "3"
},
{
"text": "He manages to tell her that she has been resurrected En: That she has been resurrected he manages to tell her Sv: Att hon har uppst\u00e5tt han lyckas ber\u00e4tta f\u00f6r henne Ru: Chto ona byla voskreshena on smog rasskazat' ej Figure 2 : Examples of the clause rotation perturbation (ClauseShift). Languages: En=English, Ru=Russian, Sv=Swedish. The English sentence is translated to the other languages for illustrational purposes.",
"cite_spans": [],
"ref_spans": [
{
"start": 216,
"end": 224,
"text": "Figure 2",
"ref_id": "FIGREF11"
}
],
"eq_spans": [],
"section": "He did not go to school yesterday",
"sec_num": null
},
{
"text": "RandomShift tests the LM sensitivity to global perturbations obtained by shuffling the word order. This type represents an extreme case of sentence permutation and is useful for comparing the behavior of the models at the scale of the perturbation complexity. An example of the randomized word order perturbation for the sentence \"She wanted to go to London\" is presented in Figure 3 . Languages: En=English, Ru=Russian, Sv=Swedish. The English sentence is translated to the other languages for illustrational purposes.",
"cite_spans": [],
"ref_spans": [
{
"start": 375,
"end": 383,
"text": "Figure 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "He did not go to school yesterday",
"sec_num": null
},
{
"text": "The experiments are run on two 12-layer multilingual Transformer models released by the Hugging-Face library (Wolf et al., 2020) :",
"cite_spans": [
{
"start": 109,
"end": 128,
"text": "(Wolf et al., 2020)",
"ref_id": "BIBREF79"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "4.1"
},
{
"text": "M-BERT 5 is pre-trained using masked language modeling (MLM) and next sentence prediction objectives, over concatenated monolingual Wikipedia corpora in 104 languages.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "4.1"
},
{
"text": "M-BART 6 is a sequence-to-sequence model that comprises a BERT encoder and an autoregressive GPT-2 decoder (Radford et al., 2019) . The model is pre-trained on the CC25 corpus in 25 languages using text infilling and sentence shuffling objectives, where it learns to predict masked word spans and reconstruct the permuted input. We use only the encoder in our experiments.",
"cite_spans": [
{
"start": 107,
"end": 129,
"text": "(Radford et al., 2019)",
"ref_id": "BIBREF52"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Models",
"sec_num": "4.1"
},
{
"text": "Parameter-free Probing We apply two unsupervised probing methods to reconstruct syntactic trees from self-attention (Self-Attention Probing) and so-called \"impact\" (Token Perturbed Masking) matrices computed by feeding the MLM models with each sentence s and its perturbed version s . The trees are induced by Chu-Liu-Edmonds algorithm (Chu, 1965; Edmonds, 1968) used to compute the Maximum Spanning Tree starting from the root of the corresponding gold dependency tree (Raganato and Tiedemann, 2018; Htut et al., 2019; Wu et al., 2020) . The probing performance is evaluated by the Undirected Unlabeled Attachment Score (UUAS), which reflects the percentage of words that have been assigned the correct head without taking the direction of relations and dependency labels into account (Klein and Manning, 2004) . Token Perturbed Masking (Wu et al., 2020) extracts global syntactic information by measuring the impact one word has on the prediction of another in an MLM. The impact matrix is similar to the self-attention matrix as it reflects the inter-word relationships in terms of Euclidean distance, except that it is derived from the outputs of the MLM head. For the sake of space, we refer the reader to Wu et al. (2020) for more details.",
"cite_spans": [
{
"start": 336,
"end": 347,
"text": "(Chu, 1965;",
"ref_id": "BIBREF9"
},
{
"start": 348,
"end": 362,
"text": "Edmonds, 1968)",
"ref_id": null
},
{
"start": 470,
"end": 500,
"text": "(Raganato and Tiedemann, 2018;",
"ref_id": "BIBREF53"
},
{
"start": 501,
"end": 519,
"text": "Htut et al., 2019;",
"ref_id": "BIBREF28"
},
{
"start": 520,
"end": 536,
"text": "Wu et al., 2020)",
"ref_id": "BIBREF80"
},
{
"start": 786,
"end": 811,
"text": "(Klein and Manning, 2004)",
"ref_id": "BIBREF33"
},
{
"start": 838,
"end": 855,
"text": "(Wu et al., 2020)",
"ref_id": "BIBREF80"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Interpretation Methods",
"sec_num": "4.2"
},
{
"text": "Representation Analysis Hessel and Schofield (2021) propose two metrics to compare contextualized representations and self-attention matrices produced by the model for each pair of sentences s and s . Token Identifiability (TI) evaluates the similarity of the LM's contextualized representations of a particular token in s and s . It is high if the token representations are similar to one another. Self-Attention Distance (SAD) measures if each token in s relates to similar words in s by computing row-wise Jensen-Shannon Divergence between the two self-attention matrices. It is low if an LM attends to the same words despite the perturbations.",
"cite_spans": [
{
"start": 24,
"end": 51,
"text": "Hessel and Schofield (2021)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Interpretation Methods",
"sec_num": "4.2"
},
{
"text": "Pseudo-perplexity Pseudo-perplexity (PPPL) is an intrinsic measure that estimates the probability of a sentence with an MLM similar to that of conventional LMs (Salazar et al., 2020) . PPPL-based measures have proved to correlate with human ratings (Lau et al., 2017), match or outperform autoregressive LMs (GPT-2) in ranking hypotheses for downstream tasks and the BLiMP benchmark (Salazar et al., 2020) , and perform at the human level in acceptability judgments (Lau et al., 2020) . We use two PPPL-based measures under implementation 7 by Lau et al. 2020to infer probabilities of the sentences and their perturbed counterparts. The MeanLP and PenLP measures are computed as the sum of pseudo-log-likelihood scores for each token in the sentence normalized by the total number of tokens. PenLP additionally scales the denominator with the exponent \u03b1 to penalize the effect of high scores.",
"cite_spans": [
{
"start": 160,
"end": 182,
"text": "(Salazar et al., 2020)",
"ref_id": "BIBREF61"
},
{
"start": 383,
"end": 405,
"text": "(Salazar et al., 2020)",
"ref_id": "BIBREF61"
},
{
"start": 466,
"end": 484,
"text": "(Lau et al., 2020)",
"ref_id": "BIBREF34"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Interpretation Methods",
"sec_num": "4.2"
},
{
"text": "Various PEs have been proposed to utilize the information about word order in the Transformer-based LMs Dufter et al., 2021) . Surprisingly, little is known about what PEs capture and how well they learn the meaning of positions. Wang and Chen (2020) among the first present an extensive study on the properties captured by PEs in different pre-trained Transformers and empirically evaluate their impact on the downstream performance for many NLP tasks. In the spirit of this work, we aim at analyzing the impact of the PEs on the syntactic probe performance. Towards this end, we consider the following three configurations of PEs of the M-BERT and M-BART models:",
"cite_spans": [
{
"start": 104,
"end": 124,
"text": "Dufter et al., 2021)",
"ref_id": "BIBREF14"
},
{
"start": 230,
"end": 250,
"text": "Wang and Chen (2020)",
"ref_id": "BIBREF75"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Positional Encoding",
"sec_num": "4.3"
},
{
"text": "(1) absolute=frozen PEs; (2) random=randomly initialized PEs; and (3) zero=zeroed PEs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Positional Encoding",
"sec_num": "4.3"
},
{
"text": "The discrepancy in the syntactic trees induced from the original sentences and their perturbed analogs is measured as the difference between the corresponding UUAS scores (\u03b4 UUAS). The lower the \u03b4 UUAS, the better is the syntax tree reconstructed from s with respect to the UUAS score for s. Figures 4 and 1 in Appendix 2 outline the task-wise heatmaps with the \u03b4 UUAS scores achieved by the M-BERT and M-BART models with absolute PEs for each layerhead pair, respectively. The models exhibit similar behavior, demonstrating positive correlation between the \u03b4 UUAS scores and the granularity of the perturbation. The overall pattern for both models is that they display little to no sensitivity to local and distant perturbations (NgramShift, ClauseShift) in contrast to the global ones (RandomShift). We provide examples of the dependency trees extracted from the self-attention matrices of the M-BERT model for the Swedish NgramShift task on Figure 5. The trees from both original (see Figure 5a ) and perturbed (see Figure 5b ) sentence versions receive the UUAS score of 0.86, demonstrating little changes in the assigned dependency heads under the local perturbation. On the contrary, randomizing word order (RandomShift) corrupts the syntactic structure significantly with a \u03b4 UUAS score of 0.33 (see Figure 8 , Appendix 2).",
"cite_spans": [],
"ref_spans": [
{
"start": 292,
"end": 307,
"text": "Figures 4 and 1",
"ref_id": "FIGREF1"
},
{
"start": 988,
"end": 997,
"text": "Figure 5a",
"ref_id": "FIGREF7"
},
{
"start": 1019,
"end": 1028,
"text": "Figure 5b",
"ref_id": "FIGREF7"
},
{
"start": 1307,
"end": 1315,
"text": "Figure 8",
"ref_id": "FIGREF14"
}
],
"eq_spans": [],
"section": "Parameter-free Probing",
"sec_num": "5.1"
},
{
"text": "Token Perturbed Masking The models show similar results to that of in Self-Attention Probing, with regards to the perturbation granularity (see Figure 6 ). In spite of that, the model performance on the NgramShift and ClauseShift reveal some differences between the encoders. M-BART generally achieves lower and close to zero \u03b4 UUAS scores, meaning to better restore the hierarchical information from the perturbed sentences (e.g., ClauseShift: [Sv, Ru]). We relate this to the fact that M-BART is pre-trained with the sentence shuffling objective.",
"cite_spans": [],
"ref_spans": [
{
"start": 144,
"end": 152,
"text": "Figure 6",
"ref_id": "FIGREF8"
}
],
"eq_spans": [],
"section": "Self-Attention Probing",
"sec_num": null
},
{
"text": "Language-wise Comparison Another observation is that there are more insensitive attention heads on the Russian tasks, possibly indicating that it is harder to distinguish from the perturbations as opposed to English and Swedish, particularly on the ClauseShift task with typically longer and syntactically more complex sentences (see Figures 4, 1, Appendix 2) . As for Swedish, which has a similar to English but stricter syntactic structure, M-BART tends to induce correct syntactic trees from the permuted sentences more frequently. This is indicated by negative \u03b4 UUAS scores on most tasks. Positional Encoding Analysis of the positional encoding shows that despite the genuine belief that positional information contributes most to syntactic structure encoding, the models do not seem to rely on it as much as might be expected. To analyze the impact of PEs from another perspective, for each pair of (s, s ) we compute the Euclidean distance (L2) between the corresponding impact (Token Perturbed Probing) and selfattention matrices (Self-Attention Probing) described in Section 4.2. The difference in the impact matrices produced by M-BERT model is generally observed only in the setting with zero PEs (see Figures 7; Figures 3-4 , Appendix 2). In contrast, there is almost no difference between the representations generated by M-BART across all configurations of the PEs (see Figures 5-7, Appendix 2) . This behavior is consistent with the head-wise results under Self-Attention Probing for all languages.",
"cite_spans": [],
"ref_spans": [
{
"start": 334,
"end": 360,
"text": "Figures 4, 1, Appendix 2)",
"ref_id": "FIGREF1"
},
{
"start": 1214,
"end": 1236,
"text": "Figures 7; Figures 3-4",
"ref_id": "FIGREF4"
},
{
"start": 1385,
"end": 1409,
"text": "Figures 5-7, Appendix 2)",
"ref_id": "FIGREF7"
}
],
"eq_spans": [],
"section": "Self-Attention Probing",
"sec_num": null
},
{
"text": "Token Identifiability The overall pattern for both models under the representation analysis is that for local and distant perturbations TI steadily decreases towards the output layer with rapid increases at layers [1, 10] (see Figure 9 , Appendix 3), and high for global perturbations (RandomShift). TI decreases when the perturbed inputs generate embeddings different from the intact ones. Despite that higher layers in both models are more sensitive, the perturbed representations remain similar to that of the original (Hessel and Schofield, 2021) .",
"cite_spans": [
{
"start": 522,
"end": 550,
"text": "(Hessel and Schofield, 2021)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [
{
"start": 227,
"end": 235,
"text": "Figure 9",
"ref_id": "FIGREF15"
}
],
"eq_spans": [],
"section": "Representation Analysis",
"sec_num": "5.2"
},
{
"text": "The results by SAD show that both models score significantly lower with random and zero PEs (see Figure 10 , Appendix 3), meaning lower sensitivity to the perturbations supported by the probing results (Section 5.1). This provides evidence that the encoders marginally rely on the positional information to induce the syntactic structure despite the distributions of the self-attention weights for the intact and perturbed sentences may differ according to the Jensen-Shannon divergence. ",
"cite_spans": [],
"ref_spans": [
{
"start": 97,
"end": 106,
"text": "Figure 10",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Self-Attention Distance",
"sec_num": null
},
{
"text": "Consistent with the results under parameter-free probing (Section 5.1) and representation analysis (Section 5.2), PPPL-based acceptability judgements 8 indicate that the encoders distinguish between the perturbations depending on their granularity. The overall trend is that for all languages the sentence pseudo-log-probability inferred from both LMs decreases with the increase of the perturbation complexity which is demonstrated by higher acceptability scores on NgramShift, but significantly lower scores on the ClauseShift and RandomShift (see Figures 11-12, Appendix 4) . The statistical significance of the PPPL distributions is confirmed with Kolmogorov-Smirnov and Wilcoxon signed-rank tests (p-value < 0.01).",
"cite_spans": [],
"ref_spans": [
{
"start": 550,
"end": 576,
"text": "Figures 11-12, Appendix 4)",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Pseudo-perplexity",
"sec_num": "5.3"
},
{
"text": "The syntactic sensitivity depends upon language At present, English remains the focal point of prior research in the field of NLP, leaving other languages understudied. Our probing experiments on the less explored languages with different word order flexibility show that M-BERT and M-BART behave slightly differently in Swedish and Russian. While M-BART better restores the corrupted syntactic structure on most of the tasks for Swedish, there are fewer attention heads sensitive to the perturbations in Russian, which is revealed through the examination of head-wise attention patterns of both models. Besides, the encoders receive lower probing performance for Russian that can be contributed to the more complex syntax and flexible word order.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "Pre-training objectives can help to improve syntactic robustness Analysis of the M-BERT and M-BART LMs that differ in the pre-training objectives shows that M-BERT achieves higher \u03b4 UUAS performance across all languages as opposed to M-BART pre-trained with the sentence shuffling objective. The lower \u03b4 UUAS probing performance indicates that M-BART better induces syntactic trees from both perturbed and intact sentences (see Section 5.1). Despite this, the representation and acceptability analysis demonstrate that M-BART is also capable of distinguishing between the perturbations (see Sections 5.2-5.3). A fruitful direction for future work is to analyze more LMs that differ in the architecture design and pre-training objectives.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "The LMs are less sensitive to more granular perturbations The results of the parameter-free probing show that M-BERT and M-BART exhibit little to no sensitivity to local perturbations within syntactic groups (NgramShift) and distant perturbations at the level of syntactic clauses (ClauseShift). In contrast, the global perturbations (RandomShift) are best distinguished by the encoders. As the granularity of the syntactic corruption increases, we observe a worse probing performance under all considered interpretation methods. Namely, the results are supported by representation analysis metrics (see Section 5.2) that indicate higher susceptibility to major changes in the sentences structure (RandomShift, ClauseShift), and the PPPL-based measures (see Section 5.3) prescribing higher acceptability scores to sentences with more granular perturbations (NgramShift). We also find that the sensitivity to the hierarchical corruption grows across layers together with the increase of the perturbation complexity, which is in line with Alleman et al. (2021) .",
"cite_spans": [
{
"start": 1037,
"end": 1058,
"text": "Alleman et al. (2021)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "M-BERT and M-BART barely use positional information to induce syntactic trees Previous research has shown that the token embeddings capture enough semantic information to restore the syntactic structure (Vilares et al., 2020; Kim et al., 2020; Rosa and Mare\u010dek, 2019) . Maudslay and Cotterell (2021) claim that syntactic abilities of BERT-based LMs are overestimated and raise the problem of isolating semantics from syntax. However, more recent studies show that Transformer encoders encode redundant information (Luo et al., 2021) , may not sufficiently capture the meaning of positions and be unimportant for downstream tasks (Wang and Chen, 2020) , including the setting with perturbed fine-tuning data (Clouatre et al., 2021) . In spirit with the latter studies, our results under different PEs configurations reveal that M-BERT and M-BART do not need the precise position information to restore the syntactic tree from their internal representations. The overall behavior is that zeroed (except for M-BERT) or even randomly initialized PEs can result in the probing performance and one with absolute positions. We suppose that despite the absolute positions of words changes during the N-gram permutation and sub-tree rotation procedures, the word order within the clauses remains almost the same as in the intact sentence .",
"cite_spans": [
{
"start": 203,
"end": 225,
"text": "(Vilares et al., 2020;",
"ref_id": "BIBREF70"
},
{
"start": 226,
"end": 243,
"text": "Kim et al., 2020;",
"ref_id": "BIBREF32"
},
{
"start": 244,
"end": 267,
"text": "Rosa and Mare\u010dek, 2019)",
"ref_id": "BIBREF56"
},
{
"start": 514,
"end": 532,
"text": "(Luo et al., 2021)",
"ref_id": "BIBREF43"
},
{
"start": 629,
"end": 650,
"text": "(Wang and Chen, 2020)",
"ref_id": "BIBREF75"
},
{
"start": 707,
"end": 730,
"text": "(Clouatre et al., 2021)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "(NgramShift, ClauseShift). That is, the more granular perturbations marginally confuse the LMs when: (i) predicting the masked word under Token Perturbation Probing which can be performed using only attention (Wang and Chen, 2020) , or (ii) judging the acceptability of the sentence where the low token pseudo-log-probability can occur at the juxtaposition of the syntactic groups, and clauses (Alleman et al., 2021) . We leave a more detailed exploration of the relationship between PEs and probing analysis for future work.",
"cite_spans": [
{
"start": 209,
"end": 230,
"text": "(Wang and Chen, 2020)",
"ref_id": "BIBREF75"
},
{
"start": 394,
"end": 416,
"text": "(Alleman et al., 2021)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "This paper presents an extension of the ongoing research on the controllable text perturbations to the multilingual setting and introspection of positional embeddings in pre-trained LMs. We introduce nine probing datasets for three Indo-European languages varying in their flexibility of the word order: English, Swedish, and Russian. The suite is constructed using language-specific heuristics carefully designed under linguistic expertise and organized by three types of syntactic perturbations: randomization of word order studied by previous research from many perspectives and less explored permutations within syntactic phrases and clauses. The method includes a combination of parameterfree probing methods based on the intermediate self-attention and contextualized representations, novel metrics for representation analysis, and acceptability judgments with pseudo-perplexity. We conduct a line of experiments to probe the syntactic sensitivity of two multilingual Transformers, M-BERT and M-BART, the latter of which is learned to reconstruct the word order during pre-training. The LMs are less sensitive to more granular pertur-bations and build hierarchical complexity towards the output layer. The analysis of the understudied relationship between the position embeddings and syntactic probe performance reveals that the position information is not necessary for inducing the hierarchical structure, which is a promising direction for a more detailed investigation. The results also show that the syntactic sensitivity may depend on the language and be enhanced by pre-training objectives. We believe there is still room for exploring the sensitivity to word order and syntactic abilities of modern LMs, specifically across a more diverse set of languages and models varying in the architecture design choices. 2 Parameter-free Probing ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "7"
},
{
"text": "https://github.com/evtaktasheva/ dependency_extraction",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "We use sentences from the CoNLL 2017 Shared Task on Multilingual Parsing from Raw Texts to Universal Dependencies(Ginter et al., 2017).3 https://wals.info",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "A brief statistics is outlined in Appendix 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Model name: bert-base-multilingual-cased. 6 Model name: facebook/mbart-large-cc25.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://github.com/jhlau/ acceptability-prediction-in-context",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "We present the results obtained by the MeanLP measure which are consistent with those of PenLP.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "Ekaterina Taktasheva and Ekaterina Artemova are partially supported by the framework of the HSE University Basic Research Program.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": null
},
{
"text": ".",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acceptability Judgements",
"sec_num": "4"
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Syntax-augmented multilingual BERT for cross-lingual transfer",
"authors": [
{
"first": "Wasi",
"middle": [],
"last": "Ahmad",
"suffix": ""
},
{
"first": "Haoran",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Kai-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Yashar",
"middle": [],
"last": "Mehdad",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing",
"volume": "1",
"issue": "",
"pages": "4538--4554",
"other_ids": {
"DOI": [
"10.18653/v1/2021.acl-long.350"
]
},
"num": null,
"urls": [],
"raw_text": "Wasi Ahmad, Haoran Li, Kai-Wei Chang, and Yashar Mehdad. 2021. Syntax-augmented multilingual BERT for cross-lingual transfer. In Proceedings of the 59th Annual Meeting of the Association for Com- putational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 4538-4554, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "On difficulties of cross-lingual transfer with order differences: A case study on dependency parsing",
"authors": [
{
"first": "Wasi",
"middle": [],
"last": "Ahmad",
"suffix": ""
},
{
"first": "Zhisong",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Xuezhe",
"middle": [],
"last": "Ma",
"suffix": ""
},
{
"first": "Eduard",
"middle": [],
"last": "Hovy",
"suffix": ""
},
{
"first": "Kai-Wei",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Nanyun",
"middle": [],
"last": "Peng",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "2440--2452",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1253"
]
},
"num": null,
"urls": [],
"raw_text": "Wasi Ahmad, Zhisong Zhang, Xuezhe Ma, Eduard Hovy, Kai-Wei Chang, and Nanyun Peng. 2019. On difficulties of cross-lingual transfer with order differ- ences: A case study on dependency parsing. In Pro- ceedings of the 2019 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technologies, Vol- ume 1 (Long and Short Papers), pages 2440-2452,",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Association for Computational Linguistics",
"authors": [
{
"first": "Minnesota",
"middle": [],
"last": "Minneapolis",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Minneapolis, Minnesota. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Syntactic perturbations reveal representational correlates of hierarchical phrase structure in pretrained language models",
"authors": [
{
"first": "Matteo",
"middle": [],
"last": "Alleman",
"suffix": ""
},
{
"first": "Jonathan",
"middle": [],
"last": "Mamou",
"suffix": ""
},
{
"first": "Miguel A Del",
"middle": [],
"last": "Rio",
"suffix": ""
},
{
"first": "Hanlin",
"middle": [],
"last": "Tang",
"suffix": ""
},
{
"first": "Yoon",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Sueyeon",
"middle": [],
"last": "Chung",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 6th Workshop on Representation Learning for NLP",
"volume": "",
"issue": "",
"pages": "263--276",
"other_ids": {
"DOI": [
"10.18653/v1/2021.repl4nlp-1.27"
]
},
"num": null,
"urls": [],
"raw_text": "Matteo Alleman, Jonathan Mamou, Miguel A Del Rio, Hanlin Tang, Yoon Kim, and SueYeon Chung. 2021. Syntactic perturbations reveal representa- tional correlates of hierarchical phrase structure in pretrained language models. In Proceedings of the 6th Workshop on Representation Learning for NLP (RepL4NLP-2021), pages 263-276, Online. Associ- ation for Computational Linguistics.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Syntax-BERT: Improving pre-trained transformers with syntax trees",
"authors": [
{
"first": "Jiangang",
"middle": [],
"last": "Bai",
"suffix": ""
},
{
"first": "Yujing",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Yiren",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Yaming",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "Jing",
"middle": [],
"last": "Bai",
"suffix": ""
},
{
"first": "Jing",
"middle": [],
"last": "Yu",
"suffix": ""
},
{
"first": "Yunhai",
"middle": [],
"last": "Tong",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
"volume": "",
"issue": "",
"pages": "3011--3020",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jiangang Bai, Yujing Wang, Yiren Chen, Yaming Yang, Jing Bai, Jing Yu, and Yunhai Tong. 2021. Syntax- BERT: Improving pre-trained transformers with syn- tax trees. In Proceedings of the 16th Conference of the European Chapter of the Association for Com- putational Linguistics: Main Volume, pages 3011- 3020, Online. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "The Syntax of Russian",
"authors": [
{
"first": "F",
"middle": [],
"last": "John",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Bailyn",
"suffix": ""
}
],
"year": 2012,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "John F Bailyn. 2012. The Syntax of Russian. Cam- bridge University Press.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Probing classifiers: Promises, shortcomings, and alternatives",
"authors": [
{
"first": "Yonatan",
"middle": [],
"last": "Belinkov",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2102.12452"
]
},
"num": null,
"urls": [],
"raw_text": "Yonatan Belinkov. 2021. Probing classifiers: Promises, shortcomings, and alternatives. arXiv preprint arXiv:2102.12452.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Analysis methods in neural language processing: A survey",
"authors": [
{
"first": "Yonatan",
"middle": [],
"last": "Belinkov",
"suffix": ""
},
{
"first": "James",
"middle": [],
"last": "Glass",
"suffix": ""
}
],
"year": 2019,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "7",
"issue": "",
"pages": "49--72",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00254"
]
},
"num": null,
"urls": [],
"raw_text": "Yonatan Belinkov and James Glass. 2019. Analysis methods in neural language processing: A survey. Transactions of the Association for Computational Linguistics, 7:49-72.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Subject and Object Positions in Swedish",
"authors": [
{
"first": "Kersti",
"middle": [],
"last": "B\u00f6rjars",
"suffix": ""
},
{
"first": "Elisabet",
"middle": [],
"last": "Engdahl",
"suffix": ""
},
{
"first": "Maia",
"middle": [],
"last": "Andr\u00e9asson",
"suffix": ""
},
{
"first": "Miriam",
"middle": [],
"last": "Butt",
"suffix": ""
},
{
"first": "Tracy Holloway",
"middle": [],
"last": "King",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the LFG03 Conference",
"volume": "",
"issue": "",
"pages": "43--58",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kersti B\u00f6rjars, Elisabet Engdahl, Maia Andr\u00e9asson, Miriam Butt, and Tracy Holloway King. 2003. Sub- ject and Object Positions in Swedish. In Proceed- ings of the LFG03 Conference, pages 43-58. Cite- seer.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "On the Shortest Arborescence of a Directed Graph",
"authors": [
{
"first": "Yoeng-Jin",
"middle": [],
"last": "Chu",
"suffix": ""
}
],
"year": 1965,
"venue": "Scientia Sinica",
"volume": "14",
"issue": "",
"pages": "1396--1400",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yoeng-Jin Chu. 1965. On the Shortest Arborescence of a Directed Graph. Scientia Sinica, 14:1396-1400.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "What does BERT look at? an analysis of BERT's attention",
"authors": [
{
"first": "Kevin",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "Urvashi",
"middle": [],
"last": "Khandelwal",
"suffix": ""
},
{
"first": "Omer",
"middle": [],
"last": "Levy",
"suffix": ""
},
{
"first": "Christopher",
"middle": [
"D"
],
"last": "Manning",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP",
"volume": "",
"issue": "",
"pages": "276--286",
"other_ids": {
"DOI": [
"10.18653/v1/W19-4828"
]
},
"num": null,
"urls": [],
"raw_text": "Kevin Clark, Urvashi Khandelwal, Omer Levy, and Christopher D. Manning. 2019. What does BERT look at? an analysis of BERT's attention. In Pro- ceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 276-286, Florence, Italy. Association for Computational Linguistics.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Amal Zouaq, and Sarath Chandar. 2021. Demystifying Neural Language Models' Insensitivity to Word-Order",
"authors": [
{
"first": "Louis",
"middle": [],
"last": "Clouatre",
"suffix": ""
},
{
"first": "Prasanna",
"middle": [],
"last": "Parthasarathi",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2107.13955"
]
},
"num": null,
"urls": [],
"raw_text": "Louis Clouatre, Prasanna Parthasarathi, Amal Zouaq, and Sarath Chandar. 2021. Demystifying Neu- ral Language Models' Insensitivity to Word-Order. arXiv preprint arXiv:2107.13955.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties",
"authors": [
{
"first": "Alexis",
"middle": [],
"last": "Conneau",
"suffix": ""
},
{
"first": "German",
"middle": [],
"last": "Kruszewski",
"suffix": ""
},
{
"first": "Guillaume",
"middle": [],
"last": "Lample",
"suffix": ""
},
{
"first": "Lo\u00efc",
"middle": [],
"last": "Barrault",
"suffix": ""
},
{
"first": "Marco",
"middle": [],
"last": "Baroni",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics",
"volume": "1",
"issue": "",
"pages": "2126--2136",
"other_ids": {
"DOI": [
"10.18653/v1/P18-1198"
]
},
"num": null,
"urls": [],
"raw_text": "Alexis Conneau, German Kruszewski, Guillaume Lam- ple, Lo\u00efc Barrault, and Marco Baroni. 2018. What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties. In Proceedings of the 56th Annual Meeting of the As- sociation for Computational Linguistics (Volume 1: Long Papers), pages 2126-2136, Melbourne, Aus- tralia. Association for Computational Linguistics.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"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": {
"DOI": [
"10.18653/v1/N19-1423"
]
},
"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 Technologies, Volume 1 (Long and Short Papers), pages 4171-4186, Minneapolis, Minnesota. Associ- ation for Computational Linguistics.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Position Information in Transformers: An Overview",
"authors": [
{
"first": "Philipp",
"middle": [],
"last": "Dufter",
"suffix": ""
},
{
"first": "Martin",
"middle": [],
"last": "Schmitt",
"suffix": ""
},
{
"first": "Hinrich",
"middle": [],
"last": "Sch\u00fctze",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2102.11090"
]
},
"num": null,
"urls": [],
"raw_text": "Philipp Dufter, Martin Schmitt, and Hinrich Sch\u00fctze. 2021. Position Information in Transformers: An Overview. arXiv preprint arXiv:2102.11090.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Optimum branchings. Mathematics and the Decision Sciences, Part",
"authors": [],
"year": 1968,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jack Edmonds. 1968. Optimum branchings. Math- ematics and the Decision Sciences, Part, 1(335- 345):25.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "How to probe sentence embeddings in low-resource languages: On structural design choices for probing task evaluation",
"authors": [
{
"first": "Steffen",
"middle": [],
"last": "Eger",
"suffix": ""
},
{
"first": "Johannes",
"middle": [],
"last": "Daxenberger",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 24th Conference on Computational Natural Language Learning",
"volume": "",
"issue": "",
"pages": "108--118",
"other_ids": {
"DOI": [
"10.18653/v1/2020.conll-1.8"
]
},
"num": null,
"urls": [],
"raw_text": "Steffen Eger, Johannes Daxenberger, and Iryna Gurevych. 2020. How to probe sentence embed- dings in low-resource languages: On structural de- sign choices for probing task evaluation. In Pro- ceedings of the 24th Conference on Computational Natural Language Learning, pages 108-118, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language models",
"authors": [
{
"first": "Allyson",
"middle": [],
"last": "Ettinger",
"suffix": ""
}
],
"year": 2020,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "8",
"issue": "",
"pages": "34--48",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00298"
]
},
"num": null,
"urls": [],
"raw_text": "Allyson Ettinger. 2020. What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language models. Transactions of the Association for Computational Linguistics, 8:34-48.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "RNNs as Psycholinguistic Subjects: Syntactic State and Grammatical Dependency",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Futrell",
"suffix": ""
},
{
"first": "Ethan",
"middle": [],
"last": "Wilcox",
"suffix": ""
},
{
"first": "Takashi",
"middle": [],
"last": "Morita",
"suffix": ""
},
{
"first": "Roger",
"middle": [],
"last": "Levy",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1809.01329"
]
},
"num": null,
"urls": [],
"raw_text": "Richard Futrell, Ethan Wilcox, Takashi Morita, and Roger Levy. 2018. RNNs as Psycholinguistic Sub- jects: Syntactic State and Grammatical Dependency. arXiv preprint arXiv:1809.01329.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Neural language models as psycholinguistic subjects: Representations of syntactic state",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Futrell",
"suffix": ""
},
{
"first": "Ethan",
"middle": [],
"last": "Wilcox",
"suffix": ""
},
{
"first": "Takashi",
"middle": [],
"last": "Morita",
"suffix": ""
},
{
"first": "Peng",
"middle": [],
"last": "Qian",
"suffix": ""
},
{
"first": "Miguel",
"middle": [],
"last": "Ballesteros",
"suffix": ""
},
{
"first": "Roger",
"middle": [],
"last": "Levy",
"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": "32--42",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1004"
]
},
"num": null,
"urls": [],
"raw_text": "Richard Futrell, Ethan Wilcox, Takashi Morita, Peng Qian, Miguel Ballesteros, and Roger Levy. 2019. Neural language models as psycholinguistic sub- jects: Representations of syntactic state. In Proceed- ings of the 2019 Conference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 32-42, Minneapolis, Minnesota. Association for Computational Linguis- tics.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "SyntaxGym: An Online Platform for Targeted Evaluation of Language Models",
"authors": [
{
"first": "Jon",
"middle": [],
"last": "Gauthier",
"suffix": ""
},
{
"first": "Jennifer",
"middle": [],
"last": "Hu",
"suffix": ""
},
{
"first": "Ethan",
"middle": [],
"last": "Wilcox",
"suffix": ""
},
{
"first": "Peng",
"middle": [],
"last": "Qian",
"suffix": ""
},
{
"first": "Roger",
"middle": [],
"last": "Levy",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations",
"volume": "",
"issue": "",
"pages": "70--76",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jon Gauthier, Jennifer Hu, Ethan Wilcox, Peng Qian, and Roger Levy. 2020. SyntaxGym: An Online Plat- form for Targeted Evaluation of Language Models. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 70-76.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "CoNLL 2017 shared task -automatically annotated raw texts and word embeddings. LINDAT/CLARIAH-CZ digital library at the Institute of Formal and Applied Linguistics (\u00daFAL), Faculty of Mathematics and Physics, Charles University",
"authors": [
{
"first": "Filip",
"middle": [],
"last": "Ginter",
"suffix": ""
},
{
"first": "Jan",
"middle": [],
"last": "Haji\u010d",
"suffix": ""
},
{
"first": "Juhani",
"middle": [],
"last": "Luotolahti",
"suffix": ""
},
{
"first": "Milan",
"middle": [],
"last": "Straka",
"suffix": ""
},
{
"first": "Daniel",
"middle": [],
"last": "Zeman",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Filip Ginter, Jan Haji\u010d, Juhani Luotolahti, Milan Straka, and Daniel Zeman. 2017. CoNLL 2017 shared task -automatically annotated raw texts and word embed- dings. LINDAT/CLARIAH-CZ digital library at the Institute of Formal and Applied Linguistics (\u00daFAL), Faculty of Mathematics and Physics, Charles Uni- versity.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Assessing BERT's syntactic abilities",
"authors": [
{
"first": "Yoav",
"middle": [],
"last": "Goldberg",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1901.05287"
]
},
"num": null,
"urls": [],
"raw_text": "Yoav Goldberg. 2019. Assessing BERT's syntactic abilities. arXiv preprint arXiv:1901.05287.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "BERT & Family Eat Word Salad: Experiments with Text Understanding",
"authors": [
{
"first": "Ashim",
"middle": [],
"last": "Gupta",
"suffix": ""
},
{
"first": "Giorgi",
"middle": [],
"last": "Kvernadze",
"suffix": ""
},
{
"first": "Vivek",
"middle": [],
"last": "Srikumar",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the AAAI Conference on Artificial Intelligence",
"volume": "35",
"issue": "",
"pages": "12946--12954",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ashim Gupta, Giorgi Kvernadze, and Vivek Srikumar. 2021. BERT & Family Eat Word Salad: Experi- ments with Text Understanding. In Proceedings of the AAAI Conference on Artificial Intelligence, vol- ume 35, pages 12946-12954.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "How effective is BERT without word ordering? implications for language understanding and data privacy",
"authors": [
{
"first": "Jack",
"middle": [],
"last": "Hessel",
"suffix": ""
},
{
"first": "Alexandra",
"middle": [],
"last": "Schofield",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing",
"volume": "2",
"issue": "",
"pages": "204--211",
"other_ids": {
"DOI": [
"10.18653/v1/2021.acl-short.27"
]
},
"num": null,
"urls": [],
"raw_text": "Jack Hessel and Alexandra Schofield. 2021. How effec- tive is BERT without word ordering? implications for language understanding and data privacy. In Pro- ceedings of the 59th Annual Meeting of the Associa- tion for Computational Linguistics and the 11th In- ternational Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 204- 211, Online. Association for Computational Linguis- tics.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Designing and interpreting probes with control tasks",
"authors": [
{
"first": "John",
"middle": [],
"last": "Hewitt",
"suffix": ""
},
{
"first": "Percy",
"middle": [],
"last": "Liang",
"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": "2733--2743",
"other_ids": {
"DOI": [
"10.18653/v1/D19-1275"
]
},
"num": null,
"urls": [],
"raw_text": "John Hewitt and Percy Liang. 2019. Designing and interpreting probes with control tasks. In Proceed- ings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th Inter- national Joint Conference on Natural Language Pro- cessing (EMNLP-IJCNLP), pages 2733-2743, Hong Kong, China. Association for Computational Lin- guistics.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "A structural probe for finding syntax in word representations",
"authors": [
{
"first": "John",
"middle": [],
"last": "Hewitt",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Christopher",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Manning",
"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": "4129--4138",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1419"
]
},
"num": null,
"urls": [],
"raw_text": "John Hewitt and Christopher D. Manning. 2019. A structural probe for finding syntax in word repre- sentations. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4129-4138, Minneapolis, Minnesota. Associ- ation for Computational Linguistics.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Learning distributed representations of sentences from unlabelled data",
"authors": [
{
"first": "Felix",
"middle": [],
"last": "Hill",
"suffix": ""
},
{
"first": "Kyunghyun",
"middle": [],
"last": "Cho",
"suffix": ""
},
{
"first": "Anna",
"middle": [],
"last": "Korhonen",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "1367--1377",
"other_ids": {
"DOI": [
"10.18653/v1/N16-1162"
]
},
"num": null,
"urls": [],
"raw_text": "Felix Hill, Kyunghyun Cho, and Anna Korhonen. 2016. Learning distributed representations of sen- tences from unlabelled data. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Hu- man Language Technologies, pages 1367-1377, San Diego, California. Association for Computational Linguistics.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Do Attention heads in BERT track syntactic dependencies? arXiv preprint",
"authors": [
{
"first": "Jason",
"middle": [],
"last": "Phu Mon Htut",
"suffix": ""
},
{
"first": "Shikha",
"middle": [],
"last": "Phang",
"suffix": ""
},
{
"first": "Samuel R",
"middle": [],
"last": "Bordia",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Bowman",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1911.12246"
]
},
"num": null,
"urls": [],
"raw_text": "Phu Mon Htut, Jason Phang, Shikha Bordia, and Samuel R Bowman. 2019. Do Attention heads in BERT track syntactic dependencies? arXiv preprint arXiv:1911.12246.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "What Does BERT Learn about the Structure of Language?",
"authors": [
{
"first": "Ganesh",
"middle": [],
"last": "Jawahar",
"suffix": ""
},
{
"first": "Beno\u00eet",
"middle": [],
"last": "Sagot",
"suffix": ""
},
{
"first": "Djam\u00e9",
"middle": [],
"last": "Seddah",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "3651--3657",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ganesh Jawahar, Beno\u00eet Sagot, and Djam\u00e9 Seddah. 2019. What Does BERT Learn about the Structure of Language? In Proceedings of the 57th Annual Meeting of the Association for Computational Lin- guistics, pages 3651-3657.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "Roles and Utilization of Attention Heads in Transformerbased Neural Language Models",
"authors": [
{
"first": "Jae-Young Jo",
"middle": [],
"last": "",
"suffix": ""
},
{
"first": "Sung-Hyon",
"middle": [],
"last": "Myaeng",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "3404--3417",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jae-young Jo and Sung-Hyon Myaeng. 2020. Roles and Utilization of Attention Heads in Transformer- based Neural Language Models. In Proceedings of the 58th Annual Meeting of the Association for Com- putational Linguistics, pages 3404-3417.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "Sharp nearby, fuzzy far away: How neural language models use context",
"authors": [
{
"first": "Urvashi",
"middle": [],
"last": "Khandelwal",
"suffix": ""
},
{
"first": "He",
"middle": [],
"last": "He",
"suffix": ""
},
{
"first": "Peng",
"middle": [],
"last": "Qi",
"suffix": ""
},
{
"first": "Dan",
"middle": [],
"last": "Jurafsky",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics",
"volume": "1",
"issue": "",
"pages": "284--294",
"other_ids": {
"DOI": [
"10.18653/v1/P18-1027"
]
},
"num": null,
"urls": [],
"raw_text": "Urvashi Khandelwal, He He, Peng Qi, and Dan Juraf- sky. 2018. Sharp nearby, fuzzy far away: How neu- ral language models use context. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 284-294, Melbourne, Australia. Association for Computational Linguistics.",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "Are pre-trained language models aware of phrases? simple but strong baselines for grammar induction",
"authors": [
{
"first": "Taeuk",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Jihun",
"middle": [],
"last": "Choi",
"suffix": ""
},
{
"first": "Daniel",
"middle": [],
"last": "Edmiston",
"suffix": ""
},
{
"first": "Sanggoo",
"middle": [],
"last": "Lee",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2002.00737"
]
},
"num": null,
"urls": [],
"raw_text": "Taeuk Kim, Jihun Choi, Daniel Edmiston, and Sang- goo Lee. 2020. Are pre-trained language mod- els aware of phrases? simple but strong base- lines for grammar induction. arXiv preprint arXiv:2002.00737.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "Corpusbased induction of syntactic structure: Models of dependency and constituency",
"authors": [
{
"first": "Dan",
"middle": [],
"last": "Klein",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Manning",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 42nd Annual Meeting of the Association for Computational Linguistics (ACL-04)",
"volume": "",
"issue": "",
"pages": "478--485",
"other_ids": {
"DOI": [
"10.3115/1218955.1219016"
]
},
"num": null,
"urls": [],
"raw_text": "Dan Klein and Christopher Manning. 2004. Corpus- based induction of syntactic structure: Models of de- pendency and constituency. In Proceedings of the 42nd Annual Meeting of the Association for Com- putational Linguistics (ACL-04), pages 478-485, Barcelona, Spain.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "How furiously can colorless green ideas sleep? sentence acceptability in context",
"authors": [
{
"first": "Carlos",
"middle": [],
"last": "Jey Han Lau",
"suffix": ""
},
{
"first": "Shalom",
"middle": [],
"last": "Armendariz",
"suffix": ""
},
{
"first": "Matthew",
"middle": [],
"last": "Lappin",
"suffix": ""
},
{
"first": "Chang",
"middle": [],
"last": "Purver",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Shu",
"suffix": ""
}
],
"year": 2020,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "8",
"issue": "",
"pages": "296--310",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00315"
]
},
"num": null,
"urls": [],
"raw_text": "Jey Han Lau, Carlos Armendariz, Shalom Lappin, Matthew Purver, and Chang Shu. 2020. How furi- ously can colorless green ideas sleep? sentence ac- ceptability in context. Transactions of the Associa- tion for Computational Linguistics, 8:296-310.",
"links": null
},
"BIBREF35": {
"ref_id": "b35",
"title": "Grammaticality, Acceptability, and Probability: A Probabilistic View of Linguistic Lnowledge",
"authors": [
{
"first": "Alexander",
"middle": [],
"last": "Jey Han Lau",
"suffix": ""
},
{
"first": "Shalom",
"middle": [],
"last": "Clark",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Lappin",
"suffix": ""
}
],
"year": 2017,
"venue": "Cognitive science",
"volume": "41",
"issue": "5",
"pages": "1202--1241",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Jey Han Lau, Alexander Clark, and Shalom Lappin. 2017. Grammaticality, Acceptability, and Probabil- ity: A Probabilistic View of Linguistic Lnowledge. Cognitive science, 41(5):1202-1241.",
"links": null
},
"BIBREF36": {
"ref_id": "b36",
"title": "BART: Denoising sequence-to-sequence pretraining for natural language generation, translation, and comprehension",
"authors": [
{
"first": "Mike",
"middle": [],
"last": "Lewis",
"suffix": ""
},
{
"first": "Yinhan",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Naman",
"middle": [],
"last": "Goyal ; Abdelrahman Mohamed",
"suffix": ""
},
{
"first": "Omer",
"middle": [],
"last": "Levy",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
},
{
"first": "Luke",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "7871--7880",
"other_ids": {
"DOI": [
"10.18653/v1/2020.acl-main.703"
]
},
"num": null,
"urls": [],
"raw_text": "Mike Lewis, Yinhan Liu, Naman Goyal, Mar- jan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre- training for natural language generation, translation, and comprehension. In Proceedings of the 58th An- nual Meeting of the Association for Computational Linguistics, pages 7871-7880, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF37": {
"ref_id": "b37",
"title": "Open sesame: Getting inside BERT's linguistic knowledge",
"authors": [
{
"first": "Yongjie",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "Yi",
"middle": [],
"last": "Chern Tan",
"suffix": ""
},
{
"first": "Robert",
"middle": [],
"last": "Frank",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP",
"volume": "",
"issue": "",
"pages": "241--253",
"other_ids": {
"DOI": [
"10.18653/v1/W19-4825"
]
},
"num": null,
"urls": [],
"raw_text": "Yongjie Lin, Yi Chern Tan, and RoBERT Frank. 2019. Open sesame: Getting inside BERT's linguistic knowledge. In Proceedings of the 2019 ACL Work- shop BlackboxNLP: Analyzing and Interpreting Neu- ral Networks for NLP, pages 241-253, Florence, Italy. Association for Computational Linguistics.",
"links": null
},
"BIBREF38": {
"ref_id": "b38",
"title": "Linguistic knowledge and transferability of contextual representations",
"authors": [
{
"first": "Nelson",
"middle": [
"F"
],
"last": "Liu",
"suffix": ""
},
{
"first": "Matt",
"middle": [],
"last": "Gardner",
"suffix": ""
},
{
"first": "Yonatan",
"middle": [],
"last": "Belinkov",
"suffix": ""
},
{
"first": "Matthew",
"middle": [
"E"
],
"last": "Peters",
"suffix": ""
},
{
"first": "Noah",
"middle": [
"A"
],
"last": "Smith",
"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": "1073--1094",
"other_ids": {
"DOI": [
"10.18653/v1/N19-1112"
]
},
"num": null,
"urls": [],
"raw_text": "Nelson F. Liu, Matt Gardner, Yonatan Belinkov, Matthew E. Peters, and Noah A. Smith. 2019a. Lin- guistic knowledge and transferability of contextual representations. In Proceedings of the 2019 Confer- ence of the North American Chapter of the Associ- ation for Computational Linguistics: Human Lan- guage Technologies, Volume 1 (Long and Short Pa- pers), pages 1073-1094, Minneapolis, Minnesota. Association for Computational Linguistics.",
"links": null
},
"BIBREF39": {
"ref_id": "b39",
"title": "2020a. A survey on contextual embeddings",
"authors": [
{
"first": "Qi",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Matt",
"suffix": ""
},
{
"first": "Phil",
"middle": [],
"last": "Kusner",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Blunsom",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2003.07278"
]
},
"num": null,
"urls": [],
"raw_text": "Qi Liu, Matt J Kusner, and Phil Blunsom. 2020a. A survey on contextual embeddings. arXiv preprint arXiv:2003.07278.",
"links": null
},
"BIBREF40": {
"ref_id": "b40",
"title": "Multilingual denoising pre-training for neural machine translation",
"authors": [
{
"first": "Yinhan",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Jiatao",
"middle": [],
"last": "Gu",
"suffix": ""
},
{
"first": "Naman",
"middle": [],
"last": "Goyal",
"suffix": ""
},
{
"first": "Xian",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Sergey",
"middle": [],
"last": "Edunov",
"suffix": ""
},
{
"first": "Marjan",
"middle": [],
"last": "Ghazvininejad",
"suffix": ""
},
{
"first": "Mike",
"middle": [],
"last": "Lewis",
"suffix": ""
},
{
"first": "Luke",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
}
],
"year": 2020,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "8",
"issue": "",
"pages": "726--742",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00343"
]
},
"num": null,
"urls": [],
"raw_text": "Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, and Luke Zettlemoyer. 2020b. Multilingual denoising pre-training for neural machine translation. Transac- tions of the Association for Computational Linguis- tics, 8:726-742.",
"links": null
},
"BIBREF41": {
"ref_id": "b41",
"title": "RoBERTa: a Robustly Optimized BERT Pre-training Approach",
"authors": [
{
"first": "Yinhan",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Myle",
"middle": [],
"last": "Ott",
"suffix": ""
},
{
"first": "Naman",
"middle": [],
"last": "Goyal",
"suffix": ""
},
{
"first": "Jingfei",
"middle": [],
"last": "Du",
"suffix": ""
},
{
"first": "Mandar",
"middle": [],
"last": "Joshi",
"suffix": ""
},
{
"first": "Danqi",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Omer",
"middle": [],
"last": "Levy",
"suffix": ""
},
{
"first": "Mike",
"middle": [],
"last": "Lewis",
"suffix": ""
},
{
"first": "Luke",
"middle": [],
"last": "Zettlemoyer",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1907.11692"
]
},
"num": null,
"urls": [],
"raw_text": "Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019b. RoBERTa: a Robustly Optimized BERT Pre-training Approach. arXiv preprint arXiv:1907.11692.",
"links": null
},
"BIBREF42": {
"ref_id": "b42",
"title": "On the Importance of Word Order Information in Cross-lingual Sequence Labeling",
"authors": [
{
"first": "Zihan",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Genta",
"suffix": ""
},
{
"first": "Samuel",
"middle": [],
"last": "Winata",
"suffix": ""
},
{
"first": "Andrea",
"middle": [],
"last": "Cahyawijaya",
"suffix": ""
},
{
"first": "Zhaojiang",
"middle": [],
"last": "Madotto",
"suffix": ""
},
{
"first": "Pascale",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Fung",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the AAAI Conference on Artificial Intelligence",
"volume": "35",
"issue": "",
"pages": "13461--13469",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zihan Liu, Genta I Winata, Samuel Cahyawijaya, An- drea Madotto, Zhaojiang Lin, and Pascale Fung. 2021. On the Importance of Word Order Informa- tion in Cross-lingual Sequence Labeling. In Pro- ceedings of the AAAI Conference on Artificial Intel- ligence, volume 35, pages 13461-13469.",
"links": null
},
"BIBREF43": {
"ref_id": "b43",
"title": "Positional artefacts propagate through masked language model embeddings",
"authors": [
{
"first": "Ziyang",
"middle": [],
"last": "Luo",
"suffix": ""
},
{
"first": "Artur",
"middle": [],
"last": "Kulmizev",
"suffix": ""
},
{
"first": "Xiaoxi",
"middle": [],
"last": "Mao",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ziyang Luo, Artur Kulmizev, and Xiaoxi Mao. 2021. Positional artefacts propagate through masked lan- guage model embeddings.",
"links": null
},
"BIBREF44": {
"ref_id": "b44",
"title": "Do Syntactic Probes Probe Syntax? Experiments with Jabberwocky Probing",
"authors": [
{
"first": "Hall",
"middle": [],
"last": "Rowan",
"suffix": ""
},
{
"first": "Ryan",
"middle": [],
"last": "Maudslay",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Cotterell",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "124--131",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rowan Hall Maudslay and Ryan Cotterell. 2021. Do Syntactic Probes Probe Syntax? Experiments with Jabberwocky Probing. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 124-131.",
"links": null
},
"BIBREF45": {
"ref_id": "b45",
"title": "Linguistic profiling of a neural language model",
"authors": [
{
"first": "Alessio",
"middle": [],
"last": "Miaschi",
"suffix": ""
},
{
"first": "Dominique",
"middle": [],
"last": "Brunato",
"suffix": ""
},
{
"first": "Felice",
"middle": [],
"last": "Dell'orletta",
"suffix": ""
},
{
"first": "Giulia",
"middle": [],
"last": "Venturi",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 28th International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "745--756",
"other_ids": {
"DOI": [
"10.18653/v1/2020.coling-main.65"
]
},
"num": null,
"urls": [],
"raw_text": "Alessio Miaschi, Dominique Brunato, Felice Dell'Orletta, and Giulia Venturi. 2020. Lin- guistic profiling of a neural language model. In Proceedings of the 28th International Conference on Computational Linguistics, pages 745-756, Barcelona, Spain (Online). International Committee on Computational Linguistics.",
"links": null
},
"BIBREF46": {
"ref_id": "b46",
"title": "Analyzing Compositionality-Sensitivity of NLI Models",
"authors": [
{
"first": "Yixin",
"middle": [],
"last": "Nie",
"suffix": ""
},
{
"first": "Yicheng",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Mohit",
"middle": [],
"last": "Bansal",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the AAAI Conference on Artificial Intelligence",
"volume": "33",
"issue": "",
"pages": "6867--6874",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yixin Nie, Yicheng Wang, and Mohit Bansal. 2019. Analyzing Compositionality-Sensitivity of NLI Models. In Proceedings of the AAAI Confer- ence on Artificial Intelligence, volume 33, pages 6867-6874.",
"links": null
},
"BIBREF47": {
"ref_id": "b47",
"title": "What Context Features Can Transformer Language Models Use?",
"authors": [
{
"first": "O'",
"middle": [],
"last": "Joe",
"suffix": ""
},
{
"first": "Jacob",
"middle": [],
"last": "Connor",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Andreas",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2106.08367"
]
},
"num": null,
"urls": [],
"raw_text": "Joe O'Connor and Jacob Andreas. 2021. What Context Features Can Transformer Language Models Use? arXiv preprint arXiv:2106.08367.",
"links": null
},
"BIBREF48": {
"ref_id": "b48",
"title": "Shuffled-token detection for refining pre-trained RoBERTa",
"authors": [
{
"first": "Subhadarshi",
"middle": [],
"last": "Panda",
"suffix": ""
},
{
"first": "Anjali",
"middle": [],
"last": "Agrawal",
"suffix": ""
},
{
"first": "Jeewon",
"middle": [],
"last": "Ha",
"suffix": ""
},
{
"first": "Benjamin",
"middle": [],
"last": "Bloch",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Student Research Workshop",
"volume": "",
"issue": "",
"pages": "88--93",
"other_ids": {
"DOI": [
"10.18653/v1/2021.naacl-srw.12"
]
},
"num": null,
"urls": [],
"raw_text": "Subhadarshi Panda, Anjali Agrawal, Jeewon Ha, and Benjamin Bloch. 2021. Shuffled-token detection for refining pre-trained RoBERTa. In Proceedings of the 2021 Conference of the North American Chap- ter of the Association for Computational Linguistics: Student Research Workshop, pages 88-93, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF49": {
"ref_id": "b49",
"title": "Scikit-learn: Machine learning in python. the",
"authors": [
{
"first": "Fabian",
"middle": [],
"last": "Pedregosa",
"suffix": ""
},
{
"first": "Ga\u00ebl",
"middle": [],
"last": "Varoquaux",
"suffix": ""
},
{
"first": "Alexandre",
"middle": [],
"last": "Gramfort",
"suffix": ""
},
{
"first": "Vincent",
"middle": [],
"last": "Michel",
"suffix": ""
},
{
"first": "Bertrand",
"middle": [],
"last": "Thirion",
"suffix": ""
},
{
"first": "Olivier",
"middle": [],
"last": "Grisel",
"suffix": ""
},
{
"first": "Mathieu",
"middle": [],
"last": "Blondel",
"suffix": ""
},
{
"first": "Peter",
"middle": [],
"last": "Prettenhofer",
"suffix": ""
},
{
"first": "Ron",
"middle": [],
"last": "Weiss",
"suffix": ""
},
{
"first": "Vincent",
"middle": [],
"last": "Dubourg",
"suffix": ""
}
],
"year": 2011,
"venue": "Journal of machine Learning research",
"volume": "12",
"issue": "",
"pages": "2825--2830",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Fabian Pedregosa, Ga\u00ebl Varoquaux, Alexandre Gram- fort, Vincent Michel, BERTrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. 2011. Scikit-learn: Machine learning in python. the Journal of machine Learning research, 12:2825-2830.",
"links": null
},
"BIBREF50": {
"ref_id": "b50",
"title": "Out of Order: How Important is the Sequential Order of Words in a Sentence in",
"authors": [
{
"first": "M",
"middle": [],
"last": "Thang",
"suffix": ""
},
{
"first": "Trung",
"middle": [],
"last": "Pham",
"suffix": ""
},
{
"first": "Long",
"middle": [],
"last": "Bui",
"suffix": ""
},
{
"first": "Anh",
"middle": [],
"last": "Mai",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Nguyen",
"suffix": ""
}
],
"year": 2020,
"venue": "Natural Language Understanding Tasks? arXiv preprint",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2012.15180"
]
},
"num": null,
"urls": [],
"raw_text": "Thang M Pham, Trung Bui, Long Mai, and Anh Nguyen. 2020. Out of Order: How Important is the Sequential Order of Words in a Sentence in Natu- ral Language Understanding Tasks? arXiv preprint arXiv:2012.15180.",
"links": null
},
"BIBREF51": {
"ref_id": "b51",
"title": "On Pragmatic Change: the Borrowing of Discourse Functions",
"authors": [
{
"first": "F",
"middle": [],
"last": "Ellen",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Prince",
"suffix": ""
}
],
"year": 1988,
"venue": "Journal of pragmatics",
"volume": "12",
"issue": "5-6",
"pages": "505--518",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ellen F Prince. 1988. On Pragmatic Change: the Bor- rowing of Discourse Functions. Journal of pragmat- ics, 12(5-6):505-518.",
"links": null
},
"BIBREF52": {
"ref_id": "b52",
"title": "Language Models are Unsupervised Multitask Learners",
"authors": [
{
"first": "Alec",
"middle": [],
"last": "Radford",
"suffix": ""
},
{
"first": "Jeff",
"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": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language Models are Unsupervised Multitask Learners.",
"links": null
},
"BIBREF53": {
"ref_id": "b53",
"title": "An analysis of encoder representations in transformerbased machine translation",
"authors": [
{
"first": "Alessandro",
"middle": [],
"last": "Raganato",
"suffix": ""
},
{
"first": "J\u00f6rg",
"middle": [],
"last": "Tiedemann",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP",
"volume": "",
"issue": "",
"pages": "287--297",
"other_ids": {
"DOI": [
"10.18653/v1/W18-5431"
]
},
"num": null,
"urls": [],
"raw_text": "Alessandro Raganato and J\u00f6rg Tiedemann. 2018. An analysis of encoder representations in transformer- based machine translation. In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 287-297, Brussels, Belgium. Association for Com- putational Linguistics.",
"links": null
},
"BIBREF54": {
"ref_id": "b54",
"title": "Probing multilingual sentence representations with X-probe",
"authors": [
{
"first": "Lilja",
"middle": [],
"last": "Vinit Ravishankar",
"suffix": ""
},
{
"first": "Erik",
"middle": [],
"last": "\u00d8vrelid",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Velldal",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 4th Workshop on Representation Learning for NLP",
"volume": "",
"issue": "",
"pages": "156--168",
"other_ids": {
"DOI": [
"10.18653/v1/W19-4318"
]
},
"num": null,
"urls": [],
"raw_text": "Vinit Ravishankar, Lilja \u00d8vrelid, and Erik Velldal. 2019. Probing multilingual sentence representations with X-probe. In Proceedings of the 4th Workshop on Representation Learning for NLP (RepL4NLP- 2019), pages 156-168, Florence, Italy. Association for Computational Linguistics.",
"links": null
},
"BIBREF55": {
"ref_id": "b55",
"title": "A primer in BERTology: What we know about how BERT works",
"authors": [
{
"first": "Anna",
"middle": [],
"last": "Rogers",
"suffix": ""
},
{
"first": "Olga",
"middle": [],
"last": "Kovaleva",
"suffix": ""
},
{
"first": "Anna",
"middle": [],
"last": "Rumshisky",
"suffix": ""
}
],
"year": 2020,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "8",
"issue": "",
"pages": "842--866",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00349"
]
},
"num": null,
"urls": [],
"raw_text": "Anna Rogers, Olga Kovaleva, and Anna Rumshisky. 2020. A primer in BERTology: What we know about how BERT works. Transactions of the Associ- ation for Computational Linguistics, 8:842-866.",
"links": null
},
"BIBREF56": {
"ref_id": "b56",
"title": "Inducing Syntactic Trees from BERT Representations",
"authors": [
{
"first": "Rudolf",
"middle": [],
"last": "Rosa",
"suffix": ""
},
{
"first": "David",
"middle": [],
"last": "Mare\u010dek",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rudolf Rosa and David Mare\u010dek. 2019. Inducing Syn- tactic Trees from BERT Representations.",
"links": null
},
"BIBREF57": {
"ref_id": "b57",
"title": "Do syntax trees help pre-trained transformers extract information?",
"authors": [
{
"first": "Devendra",
"middle": [],
"last": "Sachan",
"suffix": ""
},
{
"first": "Yuhao",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "Peng",
"middle": [],
"last": "Qi",
"suffix": ""
},
{
"first": "William",
"middle": [
"L"
],
"last": "Hamilton",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
"volume": "",
"issue": "",
"pages": "2647--2661",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Devendra Sachan, Yuhao Zhang, Peng Qi, and William L. Hamilton. 2021. Do syntax trees help pre-trained transformers extract information? In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Lin- guistics: Main Volume, pages 2647-2661, Online. Association for Computational Linguistics.",
"links": null
},
"BIBREF58": {
"ref_id": "b58",
"title": "Data augmentation via dependency tree morphing for lowresource languages",
"authors": [
{
"first": "G\u00f6zde",
"middle": [],
"last": "G\u00fcl\u015fahin",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "5004--5009",
"other_ids": {
"DOI": [
"10.18653/v1/D18-1545"
]
},
"num": null,
"urls": [],
"raw_text": "G\u00f6zde G\u00fcl\u015eahin and Mark Steedman. 2018. Data aug- mentation via dependency tree morphing for low- resource languages. In Proceedings of the 2018 Conference on Empirical Methods in Natural Lan- guage Processing, pages 5004-5009, Brussels, Bel- gium. Association for Computational Linguistics.",
"links": null
},
"BIBREF59": {
"ref_id": "b59",
"title": "Data augmentation via dependency tree morphing for low-resource languages",
"authors": [
{
"first": "G\u00f6zde",
"middle": [],
"last": "G\u00fcl\u015fahin",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Steedman",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1903.09460"
]
},
"num": null,
"urls": [],
"raw_text": "G\u00f6zde G\u00fcl\u015eahin and Mark Steedman. 2019. Data augmentation via dependency tree morphing for low-resource languages. arXiv preprint arXiv:1903.09460.",
"links": null
},
"BIBREF60": {
"ref_id": "b60",
"title": "LINSPECTOR: Multilingual probing tasks for word representations",
"authors": [
{
"first": "G\u00f6zde",
"middle": [],
"last": "G\u00fcl\u015fahin",
"suffix": ""
},
{
"first": "Clara",
"middle": [],
"last": "Vania",
"suffix": ""
},
{
"first": "Ilia",
"middle": [],
"last": "Kuznetsov",
"suffix": ""
},
{
"first": "Iryna",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2020,
"venue": "Computational Linguistics",
"volume": "46",
"issue": "2",
"pages": "335--385",
"other_ids": {
"DOI": [
"10.1162/coli_a_00376"
]
},
"num": null,
"urls": [],
"raw_text": "G\u00f6zde G\u00fcl\u015eahin, Clara Vania, Ilia Kuznetsov, and Iryna Gurevych. 2020. LINSPECTOR: Multilingual probing tasks for word representations. Computa- tional Linguistics, 46(2):335-385.",
"links": null
},
"BIBREF61": {
"ref_id": "b61",
"title": "Masked language model scoring",
"authors": [
{
"first": "Julian",
"middle": [],
"last": "Salazar",
"suffix": ""
},
{
"first": "Davis",
"middle": [],
"last": "Liang",
"suffix": ""
},
{
"first": "Toan",
"middle": [
"Q"
],
"last": "Nguyen",
"suffix": ""
},
{
"first": "Katrin",
"middle": [],
"last": "Kirchhoff",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "2699--2712",
"other_ids": {
"DOI": [
"10.18653/v1/2020.acl-main.240"
]
},
"num": null,
"urls": [],
"raw_text": "Julian Salazar, Davis Liang, Toan Q. Nguyen, and Ka- trin Kirchhoff. 2020. Masked language model scor- ing. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2699-2712, Online. Association for Compu- tational Linguistics.",
"links": null
},
"BIBREF62": {
"ref_id": "b62",
"title": "Do neural dialog systems use the conversation history effectively? an empirical study",
"authors": [
{
"first": "Chinnadhurai",
"middle": [],
"last": "Sankar",
"suffix": ""
},
{
"first": "Sandeep",
"middle": [],
"last": "Subramanian",
"suffix": ""
},
{
"first": "Christopher",
"middle": [],
"last": "Pal",
"suffix": ""
},
{
"first": "Sarath",
"middle": [],
"last": "Chandar",
"suffix": ""
},
{
"first": "Yoshua",
"middle": [],
"last": "Bengio",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "32--37",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chinnadhurai Sankar, Sandeep Subramanian, Christo- pher Pal, Sarath Chandar, and Yoshua Bengio. 2019. Do neural dialog systems use the conversation his- tory effectively? an empirical study. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 32-37.",
"links": null
},
"BIBREF63": {
"ref_id": "b63",
"title": "What does bert learn from multiplechoice reading comprehension datasets? arXiv preprint",
"authors": [
{
"first": "Chenglei",
"middle": [],
"last": "Si",
"suffix": ""
},
{
"first": "Shuohang",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Min-Yen",
"middle": [],
"last": "Kan",
"suffix": ""
},
{
"first": "Jing",
"middle": [],
"last": "Jiang",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1910.12391"
]
},
"num": null,
"urls": [],
"raw_text": "Chenglei Si, Shuohang Wang, Min-Yen Kan, and Jing Jiang. 2019. What does bert learn from multiple- choice reading comprehension datasets? arXiv preprint arXiv:1910.12391.",
"links": null
},
"BIBREF64": {
"ref_id": "b64",
"title": "Masked language modeling and the distributional hypothesis: Order word matters pre-training for little",
"authors": [
{
"first": "Koustuv",
"middle": [],
"last": "Sinha",
"suffix": ""
},
{
"first": "Robin",
"middle": [],
"last": "Jia",
"suffix": ""
},
{
"first": "Dieuwke",
"middle": [],
"last": "Hupkes",
"suffix": ""
},
{
"first": "Joelle",
"middle": [],
"last": "Pineau",
"suffix": ""
},
{
"first": "Adina",
"middle": [],
"last": "Williams",
"suffix": ""
},
{
"first": "Douwe",
"middle": [],
"last": "Kiela",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Koustuv Sinha, Robin Jia, Dieuwke Hupkes, Joelle Pineau, Adina Williams, and Douwe Kiela. 2021. Masked language modeling and the distributional hypothesis: Order word matters pre-training for lit- tle. CoRR, abs/2104.06644.",
"links": null
},
"BIBREF65": {
"ref_id": "b65",
"title": "Unnatural language inference",
"authors": [
{
"first": "Koustuv",
"middle": [],
"last": "Sinha",
"suffix": ""
},
{
"first": "Prasanna",
"middle": [],
"last": "Parthasarathi",
"suffix": ""
},
{
"first": "Joelle",
"middle": [],
"last": "Pineau",
"suffix": ""
},
{
"first": "Adina",
"middle": [],
"last": "Williams",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2101.00010"
]
},
"num": null,
"urls": [],
"raw_text": "Koustuv Sinha, Prasanna Parthasarathi, Joelle Pineau, and Adina Williams. 2020. Unnatural language in- ference. arXiv preprint arXiv:2101.00010.",
"links": null
},
"BIBREF66": {
"ref_id": "b66",
"title": "Assessing the Benchmarking Capacity of Machine Reading Comprehension Datasets",
"authors": [
{
"first": "Saku",
"middle": [],
"last": "Sugawara",
"suffix": ""
},
{
"first": "Pontus",
"middle": [],
"last": "Stenetorp",
"suffix": ""
},
{
"first": "Kentaro",
"middle": [],
"last": "Inui",
"suffix": ""
},
{
"first": "Akiko",
"middle": [],
"last": "Aizawa",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the AAAI Conference on Artificial Intelligence",
"volume": "34",
"issue": "",
"pages": "8918--8927",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Saku Sugawara, Pontus Stenetorp, Kentaro Inui, and Akiko Aizawa. 2020. Assessing the Benchmark- ing Capacity of Machine Reading Comprehension Datasets. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 8918-8927.",
"links": null
},
"BIBREF67": {
"ref_id": "b67",
"title": "Learning to organize a bag of words into sentences with neural networks: An empirical study",
"authors": [
{
"first": "Chongyang",
"middle": [],
"last": "Tao",
"suffix": ""
},
{
"first": "Shen",
"middle": [],
"last": "Gao",
"suffix": ""
},
{
"first": "Juntao",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Yansong",
"middle": [],
"last": "Feng",
"suffix": ""
},
{
"first": "Dongyan",
"middle": [],
"last": "Zhao",
"suffix": ""
},
{
"first": "Rui",
"middle": [],
"last": "Yan",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "",
"issue": "",
"pages": "1682--1691",
"other_ids": {
"DOI": [
"10.18653/v1/2021.naacl-main.134"
]
},
"num": null,
"urls": [],
"raw_text": "Chongyang Tao, Shen Gao, Juntao Li, Yansong Feng, Dongyan Zhao, and Rui Yan. 2021. Learning to or- ganize a bag of words into sentences with neural net- works: An empirical study. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Hu- man Language Technologies, pages 1682-1691, On- line. Association for Computational Linguistics.",
"links": null
},
"BIBREF68": {
"ref_id": "b68",
"title": "What do you learn from context? probing for sentence structure in contextualized word representations",
"authors": [
{
"first": "Ian",
"middle": [],
"last": "Tenney",
"suffix": ""
},
{
"first": "Patrick",
"middle": [],
"last": "Xia",
"suffix": ""
},
{
"first": "Berlin",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Adam",
"middle": [],
"last": "Poliak",
"suffix": ""
},
{
"first": "Thomas",
"middle": [],
"last": "Mccoy",
"suffix": ""
},
{
"first": "Najoung",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Benjamin",
"middle": [],
"last": "Van Durme",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Samuel",
"suffix": ""
},
{
"first": "Dipanjan",
"middle": [],
"last": "Bowman",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Das",
"suffix": ""
}
],
"year": 2018,
"venue": "International Conference on Learning Representations",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Samuel R Bowman, Dipan- jan Das, et al. 2018. What do you learn from con- text? probing for sentence structure in contextual- ized word representations. In International Confer- ence on Learning Representations.",
"links": null
},
"BIBREF69": {
"ref_id": "b69",
"title": "Attention is all you need",
"authors": [
{
"first": "Ashish",
"middle": [],
"last": "Vaswani",
"suffix": ""
},
{
"first": "Noam",
"middle": [],
"last": "Shazeer",
"suffix": ""
},
{
"first": "Niki",
"middle": [],
"last": "Parmar",
"suffix": ""
},
{
"first": "Jakob",
"middle": [],
"last": "Uszkoreit",
"suffix": ""
},
{
"first": "Llion",
"middle": [],
"last": "Jones",
"suffix": ""
},
{
"first": "Aidan",
"middle": [
"N"
],
"last": "Gomez",
"suffix": ""
},
{
"first": "\u0141ukasz",
"middle": [],
"last": "Kaiser",
"suffix": ""
},
{
"first": "Illia",
"middle": [],
"last": "Polosukhin",
"suffix": ""
}
],
"year": 2017,
"venue": "Advances in neural information processing systems",
"volume": "",
"issue": "",
"pages": "5998--6008",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, \u0141ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in neural information pro- cessing systems, pages 5998-6008.",
"links": null
},
"BIBREF70": {
"ref_id": "b70",
"title": "Parsing as Pretraining",
"authors": [
{
"first": "David",
"middle": [],
"last": "Vilares",
"suffix": ""
},
{
"first": "Michalina",
"middle": [],
"last": "Strzyz",
"suffix": ""
},
{
"first": "Anders",
"middle": [],
"last": "S\u00f8gaard",
"suffix": ""
},
{
"first": "Carlos",
"middle": [],
"last": "G\u00f3mez-Rodr\u00edguez",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "34",
"issue": "",
"pages": "9114--9121",
"other_ids": {
"DOI": [
"10.1609/aaai.v34i05.6446"
]
},
"num": null,
"urls": [],
"raw_text": "David Vilares, Michalina Strzyz, Anders S\u00f8gaard, and Carlos G\u00f3mez-Rodr\u00edguez. 2020. Parsing as Pretrain- ing. 34:9114-9121.",
"links": null
},
"BIBREF71": {
"ref_id": "b71",
"title": "DaLAJ-a Dataset for Linguistic Acceptability Judgments for Swedish: Format, Baseline",
"authors": [
{
"first": "Elena",
"middle": [],
"last": "Volodina",
"suffix": ""
},
{
"first": "Ali",
"middle": [],
"last": "Yousuf",
"suffix": ""
},
{
"first": "Julia",
"middle": [],
"last": "Mohammed",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Klezl",
"suffix": ""
}
],
"year": 2021,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:2105.06681"
]
},
"num": null,
"urls": [],
"raw_text": "Elena Volodina, Yousuf Ali Mohammed, and Julia Klezl. 2021. DaLAJ-a Dataset for Linguistic Ac- ceptability Judgments for Swedish: Format, Base- line, Sharing. arXiv preprint arXiv:2105.06681.",
"links": null
},
"BIBREF72": {
"ref_id": "b72",
"title": "GLUE: A multi-task benchmark and analysis platform for natural language understanding",
"authors": [
{
"first": "Alex",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Amanpreet",
"middle": [],
"last": "Singh",
"suffix": ""
},
{
"first": "Julian",
"middle": [],
"last": "Michael",
"suffix": ""
},
{
"first": "Felix",
"middle": [],
"last": "Hill",
"suffix": ""
},
{
"first": "Omer",
"middle": [],
"last": "Levy",
"suffix": ""
},
{
"first": "Samuel",
"middle": [],
"last": "Bowman",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 2018 EMNLP Workshop Black-boxNLP: Analyzing and Interpreting Neural Networks for NLP",
"volume": "",
"issue": "",
"pages": "353--355",
"other_ids": {
"DOI": [
"10.18653/v1/W18-5446"
]
},
"num": null,
"urls": [],
"raw_text": "Alex Wang, Amanpreet Singh, Julian Michael, Fe- lix Hill, Omer Levy, and Samuel Bowman. 2018. GLUE: A multi-task benchmark and analysis plat- form for natural language understanding. In Pro- ceedings of the 2018 EMNLP Workshop Black- boxNLP: Analyzing and Interpreting Neural Net- works for NLP, pages 353-355, Brussels, Belgium. Association for Computational Linguistics.",
"links": null
},
"BIBREF73": {
"ref_id": "b73",
"title": "On Position Embeddings in BERT",
"authors": [
{
"first": "Benyou",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Lifeng",
"middle": [],
"last": "Shang",
"suffix": ""
},
{
"first": "Christina",
"middle": [],
"last": "Lioma",
"suffix": ""
},
{
"first": "Xin",
"middle": [],
"last": "Jiang",
"suffix": ""
},
{
"first": "Hao",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "Qun",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Jakob",
"middle": [
"Grue"
],
"last": "Simonsen",
"suffix": ""
}
],
"year": 2020,
"venue": "International Conference on Learning Representations",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Benyou Wang, Lifeng Shang, Christina Lioma, Xin Jiang, Hao Yang, Qun Liu, and Jakob Grue Simon- sen. 2020. On Position Embeddings in BERT. In International Conference on Learning Representa- tions.",
"links": null
},
"BIBREF74": {
"ref_id": "b74",
"title": "Structbert: incorporating language structures into pretraining for deep language understanding",
"authors": [
{
"first": "Wei",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Bin",
"middle": [],
"last": "Bi",
"suffix": ""
},
{
"first": "Ming",
"middle": [],
"last": "Yan",
"suffix": ""
},
{
"first": "Chen",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Zuyi",
"middle": [],
"last": "Bao",
"suffix": ""
},
{
"first": "Jiangnan",
"middle": [],
"last": "Xia",
"suffix": ""
},
{
"first": "Liwei",
"middle": [],
"last": "Peng",
"suffix": ""
},
{
"first": "Luo",
"middle": [],
"last": "Si",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1908.04577"
]
},
"num": null,
"urls": [],
"raw_text": "Wei Wang, Bin Bi, Ming Yan, Chen Wu, Zuyi Bao, Jiangnan Xia, Liwei Peng, and Luo Si. 2019. Struct- bert: incorporating language structures into pre- training for deep language understanding. arXiv preprint arXiv:1908.04577.",
"links": null
},
"BIBREF75": {
"ref_id": "b75",
"title": "What do position embeddings learn? an empirical study of pre-trained language model positional encoding",
"authors": [
{
"first": "Yu-An",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Yun-Nung",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
"volume": "",
"issue": "",
"pages": "6840--6849",
"other_ids": {
"DOI": [
"10.18653/v1/2020.emnlp-main.555"
]
},
"num": null,
"urls": [],
"raw_text": "Yu-An Wang and Yun-Nung Chen. 2020. What do position embeddings learn? an empirical study of pre-trained language model positional encoding. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6840-6849, Online. Association for Computa- tional Linguistics.",
"links": null
},
"BIBREF76": {
"ref_id": "b76",
"title": "Grammatical analysis of pretrained sentence encoders with acceptability judgments",
"authors": [
{
"first": "Alex",
"middle": [],
"last": "Warstadt",
"suffix": ""
},
{
"first": "Samuel",
"middle": [
"R"
],
"last": "Bowman",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alex Warstadt and Samuel R. Bowman. 2019. Grammatical analysis of pretrained sentence en- coders with acceptability judgments. CoRR, abs/1901.03438.",
"links": null
},
"BIBREF77": {
"ref_id": "b77",
"title": "BLiMP: The benchmark of linguistic minimal pairs for English",
"authors": [
{
"first": "Alex",
"middle": [],
"last": "Warstadt",
"suffix": ""
},
{
"first": "Alicia",
"middle": [],
"last": "Parrish",
"suffix": ""
},
{
"first": "Haokun",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Anhad",
"middle": [],
"last": "Mohananey",
"suffix": ""
},
{
"first": "Wei",
"middle": [],
"last": "Peng",
"suffix": ""
},
{
"first": "Sheng-Fu",
"middle": [],
"last": "Wang",
"suffix": ""
},
{
"first": "Samuel",
"middle": [
"R"
],
"last": "Bowman",
"suffix": ""
}
],
"year": 2020,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "8",
"issue": "",
"pages": "377--392",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00321"
]
},
"num": null,
"urls": [],
"raw_text": "Alex Warstadt, Alicia Parrish, Haokun Liu, Anhad Mo- hananey, Wei Peng, Sheng-Fu Wang, and Samuel R. Bowman. 2020. BLiMP: The benchmark of linguis- tic minimal pairs for English. Transactions of the As- sociation for Computational Linguistics, 8:377-392.",
"links": null
},
"BIBREF78": {
"ref_id": "b78",
"title": "Neural network acceptability judgments",
"authors": [
{
"first": "Alex",
"middle": [],
"last": "Warstadt",
"suffix": ""
},
{
"first": "Amanpreet",
"middle": [],
"last": "Singh",
"suffix": ""
},
{
"first": "Samuel",
"middle": [
"R"
],
"last": "",
"suffix": ""
}
],
"year": 2019,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "7",
"issue": "",
"pages": "625--641",
"other_ids": {
"DOI": [
"10.1162/tacl_a_00290"
]
},
"num": null,
"urls": [],
"raw_text": "Alex Warstadt, Amanpreet Singh, and Samuel R. Bow- man. 2019. Neural network acceptability judgments. Transactions of the Association for Computational Linguistics, 7:625-641.",
"links": null
},
"BIBREF79": {
"ref_id": "b79",
"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": {
"DOI": [
"10.18653/v1/2020.emnlp-demos.6"
]
},
"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 process- ing. In Proceedings of the 2020 Conference on Em- pirical Methods in Natural Language Processing: System Demonstrations, pages 38-45, Online. Asso- ciation for Computational Linguistics.",
"links": null
},
"BIBREF80": {
"ref_id": "b80",
"title": "Perturbed masking: Parameter-free probing for analyzing and interpreting BERT",
"authors": [
{
"first": "Zhiyong",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Yun",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Ben",
"middle": [],
"last": "Kao",
"suffix": ""
},
{
"first": "Qun",
"middle": [],
"last": "Liu",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "4166--4176",
"other_ids": {
"DOI": [
"10.18653/v1/2020.acl-main.383"
]
},
"num": null,
"urls": [],
"raw_text": "Zhiyong Wu, Yun Chen, Ben Kao, and Qun Liu. 2020. Perturbed masking: Parameter-free probing for ana- lyzing and interpreting BERT. In Proceedings of the 58th Annual Meeting of the Association for Compu- tational Linguistics, pages 4166-4176, Online. As- sociation for Computational Linguistics.",
"links": null
},
"BIBREF81": {
"ref_id": "b81",
"title": "CLiMP: A benchmark for Chinese language model evaluation",
"authors": [
{
"first": "Beilei",
"middle": [],
"last": "Xiang",
"suffix": ""
},
{
"first": "Changbing",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "Yu",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Alex",
"middle": [],
"last": "Warstadt",
"suffix": ""
},
{
"first": "Katharina",
"middle": [],
"last": "Kann",
"suffix": ""
}
],
"year": 2021,
"venue": "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
"volume": "",
"issue": "",
"pages": "2784--2790",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Beilei Xiang, Changbing Yang, Yu Li, Alex Warstadt, and Katharina Kann. 2021. CLiMP: A benchmark for Chinese language model evaluation. In Proceed- ings of the 16th Conference of the European Chap- ter of the Association for Computational Linguistics: Main Volume, pages 2784-2790, Online. Associa- tion for Computational Linguistics.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "He did not go school to yesterday Ru: Vchera on ne poshel shkolu v Sv: Han gick inte skolan till ig\u00e5r"
},
"FIGREF1": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Examples of the N-gram perturbations (NgramShift). Languages: En=English, Ru=Russian, Sv=Swedish. The English sentence is translated to the other languages for illustrational purposes."
},
"FIGREF2": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "She wanted to go to London En: Wanted London go she to to Sv: Ville London \u00e5ka hon till att Ru: Hotela London poehat' ona v Figure 3: Examples of the word order shuffling (RandomShift)."
},
"FIGREF3": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Self-Attention Probing (Htut et al., 2019) allows to explore if attention heads encode complete syntactic trees. To this end, each layer-head attention matrix is treated as a weighted directed graph where the vertices represent words in the input sentence and edges are the attention weights. Modelspecific special tokens such as [CLS], [SEP], <s>, </s> are excluded at the pre-processing stage to eliminate their impact on other tokens."
},
"FIGREF4": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The task-wise heatmaps depicting the \u03b4 UUAS scores by M-BERT for each language. Method=Self-Attention Probing. PE=absolute. X-axis=Attention head index. Y-axis=Layer index. Tasks: NgramShift (top); ClauseShift (middle); RandomShift (bottom). Languages: En=English (left); Sv=Swedish (middle); Ru=Russian (right)."
},
"FIGREF5": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Figure 2(see Appendix 2) illustrates the distribution of \u03b4 UUAS scores for M-BERT with different PEs on English tasks. The heatmaps show that zero and random PEs only slightly affects the quality of the probe performance of the self-attention heads."
},
"FIGREF7": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Graphical representations of the syntactic trees inferred for the Swedish sentence Treubiaceae \u00e4r en familj av bladmossor 'Treubiaceae is a family of mosses' and its perturbed version. original=the original sentence; perturbed=the perturbed version; gold=gold standard. Task=NgramShift. Model=M-BERT (Layer: 11; Head: 2). Method=Self-Attention Probing. The perturbation is underlined with red, and incorrectly assigned dependency heads are marked with red arrows."
},
"FIGREF8": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The probing performance in \u03b4 UUAS across layers under Token Perturbed Probing. PE=absolute. The scores are averaged over attention heads at each layer. X-axis=Attention head index. Y-axis=\u03b4 UUAS."
},
"FIGREF9": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The Euclidean distances between the impact matrices computed by M-BERT with different PEs over each pair of sentences (s, s ) for Swedish. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right)"
},
"FIGREF10": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The task-wise heatmaps depicting the \u03b4 UUAS scores by M-BART for each language. Method=Self-Attention Probing. PE=absolute. X-axis=Attention head index. Y-axis=Layer index. Tasks: NgramShift (top); ClauseShift (middle); RandomShift (bottom). Languages: En=English (left); Sv=Swedish (middle); Ru=Russian (right)."
},
"FIGREF11": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The task-wise heatmaps depicting the \u03b4 UUAS scores by M-BERT for each language. Method=Self-Attention Probing. PE: absolute (left); random (middle); zero (right). X-axis=Attention head index. Y-axis=Layer index. Tasks: NgramShift (top); ClauseShift (middle); RandomShift (bottom)."
},
"FIGREF12": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The Euclidean distance between the impact matrices computed by M-BERT with different PEs over each pair of sentences (s, s ) for Russian. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right) . The Euclidean distance between the impact matrices computed by M-BERT with different PEs over each pair of sentences (s, s ) for English. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right)."
},
"FIGREF13": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "The Euclidean distance between the impact matrices computed by M-BART with different PEs over each pair of sentences (s, s ) for Swedish. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right) . The Euclidean distance between the impact matrices computed by M-BART with different PEs over each pair of sentences (s, s ) for English. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right) . The Euclidean distance between the impact matrices computed by M-BART with different PEs over each pair of sentences (s, s ) for Russian. The distances are averaged over attention heads at each layer. Method: Token Perturbed Masking. Tasks: NgramShift (left); ClauseShift (middle); RandomShift (right)"
},
"FIGREF14": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Graphical representations of the syntactic trees inferred for the English sentence Iyassu stoned me with gold and its perturbed version. original=the original sentence; perturbed=the perturbed version; gold=gold standard. Task=RandomShift. Model=M-BERT (Layer: 11; Head: 2). Method=Self-Attention Probing. The perturbation is underlined with red, and incorrectly assigned dependency heads are marked with red arrows."
},
"FIGREF15": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Token identifiability (TI) by layer for M-BERT and M-BART on the NgramShift (left) and ClauseShift (right) tasks for Russian. Dashed lines represent the scores computed over the intact sentences. X-axis=Layer index. Y-axis=TI."
},
"FIGREF16": {
"num": null,
"type_str": "figure",
"uris": null,
"text": "Self-Attention Distance (SAD) by layer for M-BART and M-BERT with absolute (left) and zeroed (right) positional embeddings on the RandomShift task for Swedish. X-axis=Layer index. Y-axis=SAD."
},
"TABREF2": {
"content": "<table/>",
"type_str": "table",
"html": null,
"num": null,
"text": "The UUAS scores by Self-Attention Probing method. The minimum and maximum values are given"
},
"TABREF3": {
"content": "<table/>",
"type_str": "table",
"html": null,
"num": null,
"text": "The UUAS scores by Token Perturbed Masking probe. The minimum and maximum values are given (min; max). Languages: Ru=Russian, En=English, Sv=Swedish."
}
}
}
} |