File size: 96,827 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
{
    "paper_id": "2020",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T13:27:12.116276Z"
    },
    "title": "Coreference for Discourse Parsing: A Neural Approach",
    "authors": [
        {
            "first": "Grigorii",
            "middle": [],
            "last": "Guz",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "University of British Columbia Vancouver",
                "location": {
                    "postCode": "V6T 1Z4",
                    "region": "BC",
                    "country": "Canada"
                }
            },
            "email": "gguz@cs.ubc.ca"
        },
        {
            "first": "Giuseppe",
            "middle": [],
            "last": "Carenini",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "University of British Columbia Vancouver",
                "location": {
                    "postCode": "V6T 1Z4",
                    "region": "BC",
                    "country": "Canada"
                }
            },
            "email": "carenini@cs.ubc.ca"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "We present preliminary results on investigating the benefits of coreference resolution features for neural RST discourse parsing by considering different levels of coupling of the discourse parser with the coreference resolver. In particular, starting with a strong baseline neural parser unaware of any coreference information, we compare a parser which utilizes only the output of a neural coreference resolver, with a more sophisticated model, where discourse parsing and coreference resolution are jointly learned in a neural multitask fashion. Results indicate that these initial attempts to incorporate coreference information do not boost the performance of discourse parsing in a statistically significant way.",
    "pdf_parse": {
        "paper_id": "2020",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "We present preliminary results on investigating the benefits of coreference resolution features for neural RST discourse parsing by considering different levels of coupling of the discourse parser with the coreference resolver. In particular, starting with a strong baseline neural parser unaware of any coreference information, we compare a parser which utilizes only the output of a neural coreference resolver, with a more sophisticated model, where discourse parsing and coreference resolution are jointly learned in a neural multitask fashion. Results indicate that these initial attempts to incorporate coreference information do not boost the performance of discourse parsing in a statistically significant way.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "Discourse parsing is a very useful Natural Language Processing (NLP) task involving predicting and analyzing discourse structures, which represent the coherence properties and relations among constituents of multi-sentential documents. In this work, we investigate discourse parsing in the context of Rhetorical Structure Theory (RST) Mann and Thompson (1988) , which encodes documents into complete constituency discourse trees. An RST tree is defined on the sequence of a document's EDUs (Elementary Discourse Units), which are clause-like sentences or sentence fragments (propositions), acting as the leaves of the tree. Adjacent EDUs and constituents are hierarchically aggregated to form (possibly non-binary) constituents, with internal nodes containing (1) a nuclearity label, defining the importance of that subtree (rooted at the internal node) in the local context and (2) a relation label, defining the type of semantic connection between the two subtrees (e.g., Elaboration, Background). (Asher and Lascarides, 2003) of a discourse being ill-formed due to the invalid anaphoric link. Previous research has shown that the use of RSTstyle discourse parsing as a system component can enhance important tasks, such as sentiment analysis, summarization and text categorization (Bhatia et al., 2015; Nejat et al., 2017; Hogenboom et al., 2015; Gerani et al., 2014; Ji and Smith, 2017) . And more recently, it has been found that RST discourse structures can complement learned contextual embeddings (e.g., BERT (Devlin et al., 2018) ), in tasks where linguistic information on complete documents is critical, such as argumentation analysis (Chakrabarty et al., 2019) .",
                "cite_spans": [
                    {
                        "start": 353,
                        "end": 359,
                        "text": "(1988)",
                        "ref_id": null
                    },
                    {
                        "start": 1000,
                        "end": 1028,
                        "text": "(Asher and Lascarides, 2003)",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 1284,
                        "end": 1305,
                        "text": "(Bhatia et al., 2015;",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 1306,
                        "end": 1325,
                        "text": "Nejat et al., 2017;",
                        "ref_id": "BIBREF26"
                    },
                    {
                        "start": 1326,
                        "end": 1349,
                        "text": "Hogenboom et al., 2015;",
                        "ref_id": "BIBREF13"
                    },
                    {
                        "start": 1350,
                        "end": 1370,
                        "text": "Gerani et al., 2014;",
                        "ref_id": "BIBREF10"
                    },
                    {
                        "start": 1371,
                        "end": 1390,
                        "text": "Ji and Smith, 2017)",
                        "ref_id": "BIBREF16"
                    },
                    {
                        "start": 1517,
                        "end": 1538,
                        "text": "(Devlin et al., 2018)",
                        "ref_id": "BIBREF8"
                    },
                    {
                        "start": 1646,
                        "end": 1672,
                        "text": "(Chakrabarty et al., 2019)",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction and Task Description",
                "sec_num": "1"
            },
            {
                "text": "In this work, we present preliminary results of investigating the benefits of coreference resolution features for RST discourse parsing. From the theoretical perspective, it has long been established (Asher and Lascarides, 2003) that discourse structure can impose constraints on mention antecedent distributions, with these constraints being derived from the role of each discourse unit (sen-tence or EDU) with respect to the global discourse. The Veins theory (Cristea et al., 1998) is the most known formalization of anaphoric constraints with respect to RST tree structures, involving assigning to each EDU a subset of preceding EDUs defined by the nuclearity attributes of the EDU's parent nodes in the document's discourse tree (see Appendix A for the exact definition). These constrains act as a domain of referential accessibility where the antecedents must reside, for otherwise the discourse would be considered incoherent. As an example of this phenomenon, consider the discourse structure in Figure 1 . In principle, a reader could apply commonsense knowledge to resolve the pronoun it in the last sentence to salmon in the third sentence, any proficient English speaker would call such a discourse ill-formed and incoherent, due to the fact that it breaks the discourse-imposed antecedent scope. In general, anaphora can only be resolved with respect to the most salient (sentence 1 in Figure 1 ) units of the preceding discourse (Asher and Lascarides, 2003) . For our purposes, this means that having access to a document's coreference structure might be beneficial to the task of predicting the discourse structure, since the coreference structure can constrain the discourse parser's solution space. However, as shown in a corpus study by Zeldes (2017) , the antecedent boundaries defined by Veins Theory are often too restrictive, suggesting that while discourse structures can be useful for predicting coreference structures and vice versa, these mutual constrains must be defined softly, at least in the context of RST theory.",
                "cite_spans": [
                    {
                        "start": 200,
                        "end": 228,
                        "text": "(Asher and Lascarides, 2003)",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 462,
                        "end": 484,
                        "text": "(Cristea et al., 1998)",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 1443,
                        "end": 1471,
                        "text": "(Asher and Lascarides, 2003)",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 1755,
                        "end": 1768,
                        "text": "Zeldes (2017)",
                        "ref_id": "BIBREF33"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 1004,
                        "end": 1012,
                        "text": "Figure 1",
                        "ref_id": "FIGREF0"
                    },
                    {
                        "start": 1399,
                        "end": 1407,
                        "text": "Figure 1",
                        "ref_id": "FIGREF0"
                    }
                ],
                "eq_spans": [],
                "section": "Introduction and Task Description",
                "sec_num": "1"
            },
            {
                "text": "To explore these ideas computationally with respect to modern neural models, we investigate the utility of automatically extracted coreference features and discourse-coreference shared representations in the context and for the benefit of neural RST discourse parsing. Our strong baseline SpanBERT-NoCoref utilizes SpanBERT (Joshi et al., 2020) as in the current SOTA coreference resolver, without utilizing any direct coreference information. Next, our SpanBERT-CorefFeats considers the output of coreference resolver as per Dai and Huang (2019) , letting us test the benefit of predicted and so possibly noisy coreference features. Finally, our more sophisticated SpanBERT-Multitask model learns discourse parsing together with coreference resolution in the neural multitask learning fashion, sharing the SpanBERT contextual word encoder for both models.",
                "cite_spans": [
                    {
                        "start": 324,
                        "end": 344,
                        "text": "(Joshi et al., 2020)",
                        "ref_id": "BIBREF17"
                    },
                    {
                        "start": 526,
                        "end": 546,
                        "text": "Dai and Huang (2019)",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction and Task Description",
                "sec_num": "1"
            },
            {
                "text": "Dai and Huang (2019) have already explored the benefit of using coreference information for neural PDTB implicit discourse relation classification, in a way similar to our SpanBERT-CorefFeats model. In our study, we also explore the use of shared encoder architecture for both tasks to detect the additional possible synergy.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "Modelwise, the most common approach to infer discourse trees is the linear bottom-up shiftreduce method, adopted from syntactic parsing. Wang et al. (2017) uses hand-crafted features and the shift-reduce method predicted by two separate Support-Vector-Machines (SVMs) for structureand nuclearity-prediction and relation-estimation. The neural model by Yu et al. (2018) uses a similar topology, but instead relies entirely on LSTMs for automatic feature extraction and on a single multilayer-perceptron (MLP) for classifying all possible actions. Top-down approaches to discourse parsing are also quite promising, with recent work of Kobayashi et al. (2020) applying ELMO (Peters et al., 2018) for computing span representations and achieving the new absolute SOTA performance, reporting however the scores of an ensemble of five independent runs of their proposed model instead of single-model results. In this work we follow the shift-reduce strategy and apply SpanBERT-Base (Joshi et al., 2020; Wolf et al., 2020 ), which we introduce below, for encoding the document contents.",
                "cite_spans": [
                    {
                        "start": 137,
                        "end": 155,
                        "text": "Wang et al. (2017)",
                        "ref_id": "BIBREF30"
                    },
                    {
                        "start": 352,
                        "end": 368,
                        "text": "Yu et al. (2018)",
                        "ref_id": "BIBREF32"
                    },
                    {
                        "start": 633,
                        "end": 656,
                        "text": "Kobayashi et al. (2020)",
                        "ref_id": "BIBREF19"
                    },
                    {
                        "start": 671,
                        "end": 692,
                        "text": "(Peters et al., 2018)",
                        "ref_id": "BIBREF28"
                    },
                    {
                        "start": 976,
                        "end": 996,
                        "text": "(Joshi et al., 2020;",
                        "ref_id": "BIBREF17"
                    },
                    {
                        "start": 997,
                        "end": 1014,
                        "text": "Wolf et al., 2020",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "The field of coreference resolution has recently been dominated by deep learning models. The current SOTA model by Joshi et al. (2020) is built upon the neural coreference resolver of by incorporating SpanBERT language model, which modifies the commonly used BERT (Devlin et al., 2019) architecture with a novel span masking pretraining objective. In our work, we reimplemented their coreference resolver in PyTorch (Paszke et al., 2019) . Our code for both models is available 1 .",
                "cite_spans": [
                    {
                        "start": 115,
                        "end": 134,
                        "text": "Joshi et al. (2020)",
                        "ref_id": "BIBREF17"
                    },
                    {
                        "start": 264,
                        "end": 285,
                        "text": "(Devlin et al., 2019)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 416,
                        "end": 437,
                        "text": "(Paszke et al., 2019)",
                        "ref_id": "BIBREF27"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "All our proposed parsers share the same basic shiftreduce architecture, consisting of a Queue, which is initially filled with documents EDUs in order from first to last one, and a Stack, which is initially empty, as well as the following actions on them:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Shift-Reduce Architecture",
                "sec_num": "3"
            },
            {
                "text": "w i ... w i+1 w i+2 ... w j w j+1 ... w i+1 ... w i ... w j w j+1 ... s i ... s i+1 s i+2 ... s j s j+1 ... s i ... s i+1 s i+2 ... s j s j+1 ... SpanBERT v S2 v S1 w i+2 v S2 v S1 w ' i+1 w ' i ... w ' j w ' j+1 w ' i+2 ... ... ... ... S 2 S 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Shift-Reduce Architecture",
                "sec_num": "3"
            },
            {
                "text": "The Shift delays aggregations of sub-trees at the beginning of the document by popping the top EDU Q 1 on the queue and pushing it onto the stack.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Shift-Reduce Architecture",
                "sec_num": "3"
            },
            {
                "text": "The Reduce-X aggregates the top subtrees (S 1 , S 2 ) on the stack into a single subtree (S 1\u22122 ). Each reduce action further defines a nuclearity assignment X N \u2208 {NN, NS, SN} to the nodes covered by S 1\u22122 and a relation X R \u2208 {Elaboration, Contrast, ...} holding between them.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Shift-Reduce Architecture",
                "sec_num": "3"
            },
            {
                "text": "Similarly to (Wang et al., 2017; Yu et al., 2018) , all models under consideration utilize the information from top two elements S 1 , S 2 of the stack, and top element Q 1 of the queue. In addition to word-/word+coference-based representations",
                "cite_spans": [
                    {
                        "start": 13,
                        "end": 32,
                        "text": "(Wang et al., 2017;",
                        "ref_id": "BIBREF30"
                    },
                    {
                        "start": 33,
                        "end": 49,
                        "text": "Yu et al., 2018)",
                        "ref_id": "BIBREF32"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Action Classifier Parametrization",
                "sec_num": "3.1"
            },
            {
                "text": "v S 2 , v S 1 , v Q 1",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Action Classifier Parametrization",
                "sec_num": "3.1"
            },
            {
                "text": "for these nodes, computed differently by each model as described below, we extract textual organization features of Wang et al. (2017) . In particular, for each pair S 2 \u2212 S 1 and S 1 \u2212 Q 1 , we extract indicator features representing whether the pair is within the same sentence or paragraph; for each of S 2 , S 1 and Q 1 we compute whether each of them are at the start/end of a sentence/paragraph/document.",
                "cite_spans": [
                    {
                        "start": 116,
                        "end": 134,
                        "text": "Wang et al. (2017)",
                        "ref_id": "BIBREF30"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Action Classifier Parametrization",
                "sec_num": "3.1"
            },
            {
                "text": "In accord with Wang et al. (2017) , the parsing action at each timestep is chosen by two trainable classifiers, being multi-layer perceptrons (MLPs) in our system, where each classifier takes in the concatenation of v S 2 , v S 1 , v Q 1 , together with the dense embeddings for the aforementioned organizational features. The first classifer predicts the action and nuclearity assignment among y Act,N uc \u2208 {Shif t, Reduce N N , Reduce N S , Reduce SN }, and in case the Reduce action is chosen, the second classifier predicts the discourse relation among 18 coarse-grained RST relation classes y Rel \u2208 {Attribution, Elaboration, ...}.",
                "cite_spans": [
                    {
                        "start": 15,
                        "end": 33,
                        "text": "Wang et al. (2017)",
                        "ref_id": "BIBREF30"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Action Classifier Parametrization",
                "sec_num": "3.1"
            },
            {
                "text": "Both classifiers are trained using the Cross-Entropy loss, computed for each Stack-Queue parsing step. At test time, we apply the greedy decoding strategy to predict the discourse structure.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Action Classifier Training and Inference",
                "sec_num": "3.2"
            },
            {
                "text": "We now describe the three proposed discourse parsing models which differ in the levels of coupling with the coreference model. See Figure 2 for the visual comparison.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 131,
                        "end": 139,
                        "text": "Figure 2",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "SpanBERT-NoCoref: in addition to the organizational features, our baseline system utilizes only the output SpanBERT-contextualized word embeddings. To predict each Stack-Queue action, a full document is passed through SpanBERT in a non-overlapping sliding window fashion, as per Joshi et al. (2020) , so that the context of full document can be considered for each parsing action to account for possible context-sensitivity of discourse structures (Dai and Huang, 2018) . The node representation v Q 1 for the first Queue element is computed as the mean of the first and the last word embedding of the EDU that this Queue element represents. v S 1 and v S 2 are computed as the means of the first and the last word embeddings of the nuclear EDU of S 1 and S 2 , as each non-leaf node in an RST structure encodes a relation between nuclear EDUs of its children (Morey et al., 2018) .",
                "cite_spans": [
                    {
                        "start": 279,
                        "end": 298,
                        "text": "Joshi et al. (2020)",
                        "ref_id": "BIBREF17"
                    },
                    {
                        "start": 448,
                        "end": 469,
                        "text": "(Dai and Huang, 2018)",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 860,
                        "end": 880,
                        "text": "(Morey et al., 2018)",
                        "ref_id": "BIBREF25"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "SpanBERT-CorefFeats: with this architecture variant, we attempt to assess the benefit of coreference features generated by the coreference resolver for RST parsing. Given a document with n words, the coreference features will be used to update the initial (not contextualized) SpanBERT word embeddings w 1:n , which will later be passed to SpanBERT.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "Specifically, for a given document we apply the pre-trained coreference parser of Joshi et al. (2020) to extract the document's coreference clusters C 1 , C 2 , ..., each of which are equivalence classes representing different mentions of the same entity. Afterwards, we compute the vector representation c i for each cluster C i by performing attention-based averaging over word-vectors corresponding to mentions in that cluster:",
                "cite_spans": [
                    {
                        "start": 82,
                        "end": 101,
                        "text": "Joshi et al. (2020)",
                        "ref_id": "BIBREF17"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "c i = k\u2208C i a k w k",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "where w k \u2208 R d is the initial SpanBERT word embedding for word k and a k \u2208 [0, 1] are attention scores. These cluster representations are then used for updating the document's word representations using the gating mechanism : for each word w k \u2208 D,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "f k = \u03c3(W [c i ; w k ]) w k = f k \u2022 w k + (1 \u2212 f k ) \u2022 c i if w k \u2208 C i w k otherwise",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "Finally, the embeddings w k are passed to Span-BERT for contextualization, and the node rep-",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "resentations v S 1 , v S 2 , v Q 1 are computed as in SpanBERT-NoCoref.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "SpanBERT-MultiTask: learns discourse parsing and coreference resolution in a multitask learning regime, weight-sharing the SpanBERT encoder module. The coreference resolver training step proceeds in the same fashion as in (Joshi et al., 2020) . For updating the discourse parsing model, we use the pre-computed coreference clusters C i obtained from the pretrained coreference model, as running it at every training step was prohibitively time-consuming. Using the contextualized Span-BERT word embeddings s 1:n for all words in the document, we check these coreference clusters for overlaps: considering a pair of spans S 1 , S 2 , if a cluster C i has entity mentions in the spans of both stack elements S 1 and S 2 , so that if there are mentions m j , m k \u2208 C i such that m j \u2208 S 1 and m k \u2208 S 2 , we update the span representation v S 1 (computed as in SpanBERT-NoCoref) with the attention weighted sum of mentions m k \u2208 C i \u2229S 2 by applying the gating mechanism as in SpanBERT-CorefFeats, so that the span representation for S 1 can incorporate more relevant context from S 2 . The representation for v S 2 is computed similarly using mentions m k \u2208 C i \u2229 S 1 , and the analogous computation is performed for S 1 \u2212 Q 1 pair.",
                "cite_spans": [
                    {
                        "start": 222,
                        "end": 242,
                        "text": "(Joshi et al., 2020)",
                        "ref_id": "BIBREF17"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "For learning both tasks at the same time, we utilize the approach similar to (Sanh et al., 2018) , where gradient updates are performed separately for each task and the probability of sampling a task is proportional to the relative size of each task's dataset. The initial shared SpanBERT encoder weights are set from the pretrained coreference resolver checkpoint. ",
                "cite_spans": [
                    {
                        "start": 77,
                        "end": 96,
                        "text": "(Sanh et al., 2018)",
                        "ref_id": "BIBREF29"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Proposed Models",
                "sec_num": "4"
            },
            {
                "text": "All models were trained on the RST-DT (Carlson et al., 2002) and evaluated with RST-Parseval procedure (Marcu, 2000) , with the coreference component of SpanBERT-MultiTask being trained on full OntoNotes 5.0 corpus (Weischedel et al., 2013) . The details of the training procedure such as hyperparameter assignment are outlined in the Appendix B. The test results are presented in Table 1 and are the average and standard deviation single-model scores of five independent runs. Firstly, we observe that our models strongly outperform all previous approaches, indicating huge benefit of pretrained language models for RST discourse parsing, with results approaching human performance. Then, with respect to coreference features, we notice that the models utilizing coreference information are statistically equivalent in performance to the SpanBERT-NoCoref baseline, while displaying higher variance of the test scores for Nuclearity and Relation prediction. This suggests four plausible (and not mutually exclusive) explanations: (1) the coreference information relevant to discourse parsing is already captured by SpanBERT, (2) or that coreference information is not a strong signal for discourse structure (Zeldes, 2017) , or that (3) the coreference information extracted automatically is too noisy, or that (4) our specific ways of combining coreference with discourse parsing are not adequate and more work is needed to develop better solutions. It should also be noted that we only experimented on a single discourse parsing dataset, so the conclusions or generalizations should be considered preliminary.",
                "cite_spans": [
                    {
                        "start": 38,
                        "end": 60,
                        "text": "(Carlson et al., 2002)",
                        "ref_id": "BIBREF3"
                    },
                    {
                        "start": 103,
                        "end": 116,
                        "text": "(Marcu, 2000)",
                        "ref_id": "BIBREF23"
                    },
                    {
                        "start": 215,
                        "end": 240,
                        "text": "(Weischedel et al., 2013)",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 1208,
                        "end": 1222,
                        "text": "(Zeldes, 2017)",
                        "ref_id": "BIBREF33"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 381,
                        "end": 388,
                        "text": "Table 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Experimental Settings and Results",
                "sec_num": "5"
            },
            {
                "text": "In an attempt to shed some light on the results, we compare the predicted and gold subtree from one of the documents in our development set on Figure 3 . The trees were analyzed using the RST tree visualization tool by Huber (2019) . According to Veins theory, the pronoun [neither] in EDU 8 is a mention that should have access to its mentions ([Finnair and SAS] or [they]) in preceding EDUs. However, accoding to the discourse structure predicted by SpanBERT-NoCoref, the vein for node (8) does not contain the EDUs (6) and 7(and in fact any of its preceding EDUs), so that [neither] cannot be linked to any of its preceding mentions. On the other hand, according to the gold discourse structure, EDU 8 has EDU 7 on its vein, meaning that this anaphora can be resolved. This means that if one had access to gold coreference structure and applied Veins Theory strictly, the substructure produced by SpanBERT-NoCoref would not be permitted.",
                "cite_spans": [
                    {
                        "start": 219,
                        "end": 231,
                        "text": "Huber (2019)",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 273,
                        "end": 282,
                        "text": "[neither]",
                        "ref_id": null
                    }
                ],
                "ref_spans": [
                    {
                        "start": 143,
                        "end": 151,
                        "text": "Figure 3",
                        "ref_id": "FIGREF2"
                    }
                ],
                "eq_spans": [],
                "section": "Experimental Settings and Results",
                "sec_num": "5"
            },
            {
                "text": "We empirically compare different levels of coupling between a shift-reduce neural discourse parser and a neural coreference resolver. Remarkably, our baseline delivers SOTA performance on RST-DT, but does not seem to benefit from coreference features. For future work, we plan to experiment with (1) alternative discourse parsing architectures and approaches for neural multitasking, along with more powerful coreference models (2) alternative ways of augmenting a neural discourse parser with coreference information and other tasks like summarization (3) improving the coreference resolution performance by leveraging information provided by a discourse parser.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions and Future Work",
                "sec_num": "6"
            },
            {
                "text": "http://www.cs.ubc.ca/ cs-research/lci/research-groups/ natural-language-processing/index.html",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [
            {
                "text": "The following definitions are from Cristea et al. (1998) . For each node in an RST tree, its head is defined as follows:1. The head of the terminal (leaf) node is itself.2. The head of a non-terminal node is the concatenation of the heads of its nuclear children.Next, we define the vein expression of each node recursively top-down. When the node is a leaf, the preceding nodes on its vein correspond to its domain of referential accessibility.1. The vein expression of the root is its head.2. For each nuclear node, its vein expression is the union of its head with:\u2022 its parent's vein, if this node has no left siblings that are satellites. \u2022 its parent's vein and its left sibling's vein, if this sibling is a satellite.3. For every satellite node, its vein expression is the union of its head with:\u2022 its parent's vein, if this node is a left child. \u2022 its parent's vein with heads of prior (up in the tree) satellite nodes removed.",
                "cite_spans": [
                    {
                        "start": 35,
                        "end": 56,
                        "text": "Cristea et al. (1998)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "A Veins Theory Definitions",
                "sec_num": null
            },
            {
                "text": "As RST-DT does not specify a standard trainingvalidation split, we select 10% of the training documents for the validation set, stratifying the split by the number of EDUs in each document. Similarly to Joshi et al. 2020, we train all of our models with AdamW (Loshchilov and Hutter, 2019) optimizer with learning rate of 1e \u22125 for SpanBERT and 2e \u22124 for model-specific components, with the batch size of 5 and linear decay for 20 epochs. All of our MLPs consist of 2 linear layers, with a GeLU (Hendrycks and Gimpel, 2016) nonlinearity and a Dropout layer with a value of 0.3 between them. Each organizational feature of Wang et al. (2017) is represented using a learnable 10-dimentional embedding, or a vector of zeros if the feature is missing (for example, the feature specifying if the 2-top elements of the stack are in the same sentence when the stack contains only one element). With regards to multitask regime, the probability of discourse parsing task being sampled over coreference resolution was \u2248 0.72 (each Stack-Queue state was treated as a datapoint), but due to highly demanding computational requirements of the coreference resolver and time constrains, this probability was increased to 0.9. Nonetheless, the results for the correct task proportions will be provided through other sources.",
                "cite_spans": [
                    {
                        "start": 495,
                        "end": 523,
                        "text": "(Hendrycks and Gimpel, 2016)",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 622,
                        "end": 640,
                        "text": "Wang et al. (2017)",
                        "ref_id": "BIBREF30"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "B Hyperparameters and Training Settings",
                "sec_num": null
            }
        ],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Ontonotes release 5.0. Linguistic Data Consortium",
                "authors": [
                    {
                        "first": "Ralph",
                        "middle": [],
                        "last": "Weischedel",
                        "suffix": ""
                    }
                ],
                "year": 2013,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ralph Weischedel et al. 2013. Ontonotes release 5.0. Linguistic Data Consortium.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "Logics of Conversation",
                "authors": [
                    {
                        "first": "Nicholas",
                        "middle": [],
                        "last": "Asher",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Lascarides",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Nicholas Asher and Alex Lascarides. 2003. Logics of Conversation. Cambridge University Press.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Better document-level sentiment analysis from rst discourse parsing",
                "authors": [
                    {
                        "first": "Parminder",
                        "middle": [],
                        "last": "Bhatia",
                        "suffix": ""
                    },
                    {
                        "first": "Yangfeng",
                        "middle": [],
                        "last": "Ji",
                        "suffix": ""
                    },
                    {
                        "first": "Jacob",
                        "middle": [],
                        "last": "Eisenstein",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "2212--2218",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Parminder Bhatia, Yangfeng Ji, and Jacob Eisenstein. 2015. Better document-level sentiment analysis from rst discourse parsing. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2212-2218.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "RST discourse treebank. Linguistic Data Consortium",
                "authors": [
                    {
                        "first": "Lynn",
                        "middle": [],
                        "last": "Carlson",
                        "suffix": ""
                    },
                    {
                        "first": "Mary",
                        "middle": [
                            "Ellen"
                        ],
                        "last": "Okurowski",
                        "suffix": ""
                    },
                    {
                        "first": "Daniel",
                        "middle": [],
                        "last": "Marcu",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lynn Carlson, Mary Ellen Okurowski, and Daniel Marcu. 2002. RST discourse treebank. Linguistic Data Consortium, University of Pennsylvania.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Ampersand: Argument mining for persuasive online discussions",
                "authors": [
                    {
                        "first": "Tuhin",
                        "middle": [],
                        "last": "Chakrabarty",
                        "suffix": ""
                    },
                    {
                        "first": "Christopher",
                        "middle": [],
                        "last": "Hidey",
                        "suffix": ""
                    },
                    {
                        "first": "Smaranda",
                        "middle": [],
                        "last": "Muresan",
                        "suffix": ""
                    },
                    {
                        "first": "Kathleen",
                        "middle": [],
                        "last": "Mckeown",
                        "suffix": ""
                    },
                    {
                        "first": "Alyssa",
                        "middle": [],
                        "last": "Hwang",
                        "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": "2926--2936",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Tuhin Chakrabarty, Christopher Hidey, Smaranda Muresan, Kathleen McKeown, and Alyssa Hwang. 2019. Ampersand: Argument mining for persuasive online discussions. In Proceedings of the 2019 Con- ference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP- IJCNLP), pages 2926-2936.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Veins theory: A model of global discourse cohesion and coherence",
                "authors": [
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Cristea",
                        "suffix": ""
                    },
                    {
                        "first": "Nancy",
                        "middle": [],
                        "last": "Ide",
                        "suffix": ""
                    },
                    {
                        "first": "Laurent",
                        "middle": [],
                        "last": "Romary",
                        "suffix": ""
                    }
                ],
                "year": 1998,
                "venue": "36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "281--285",
                "other_ids": {
                    "DOI": [
                        "10.3115/980845.980891"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Dan Cristea, Nancy Ide, and Laurent Romary. 1998. Veins theory: A model of global discourse cohesion and coherence. In 36th Annual Meeting of the Asso- ciation for Computational Linguistics and 17th In- ternational Conference on Computational Linguis- tics, Volume 1, pages 281-285, Montreal, Quebec, Canada. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Improving implicit discourse relation classification by modeling inter-dependencies of discourse units in a paragraph",
                "authors": [
                    {
                        "first": "Zeyu",
                        "middle": [],
                        "last": "Dai",
                        "suffix": ""
                    },
                    {
                        "first": "Ruihong",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
                "volume": "1",
                "issue": "",
                "pages": "141--151",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/N18-1013"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Zeyu Dai and Ruihong Huang. 2018. Improving im- plicit discourse relation classification by modeling inter-dependencies of discourse units in a paragraph. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Compu- tational Linguistics: Human Language Technolo- gies, Volume 1 (Long Papers), pages 141-151, New Orleans, Louisiana. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "A regularization approach for incorporating event knowledge and coreference relations into neural discourse parsing",
                "authors": [
                    {
                        "first": "Zeyu",
                        "middle": [],
                        "last": "Dai",
                        "suffix": ""
                    },
                    {
                        "first": "Ruihong",
                        "middle": [],
                        "last": "Huang",
                        "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": "2976--2987",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/D19-1295"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Zeyu Dai and Ruihong Huang. 2019. A regularization approach for incorporating event knowledge and coreference relations into neural discourse parsing. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Lan- guage Processing (EMNLP-IJCNLP), pages 2976- 2987, Hong Kong, China. Association for Computa- tional Linguistics.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Bert: Pre-training of deep bidirectional transformers for language understanding",
                "authors": [
                    {
                        "first": "Jacob",
                        "middle": [],
                        "last": "Devlin",
                        "suffix": ""
                    },
                    {
                        "first": "Ming-Wei",
                        "middle": [],
                        "last": "Chang",
                        "suffix": ""
                    },
                    {
                        "first": "Kenton",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "Kristina",
                        "middle": [],
                        "last": "Toutanova",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1810.04805"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understand- ing. arXiv preprint arXiv:1810.04805.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "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
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Abstractive summarization of product reviews using discourse structure",
                "authors": [
                    {
                        "first": "Shima",
                        "middle": [],
                        "last": "Gerani",
                        "suffix": ""
                    },
                    {
                        "first": "Yashar",
                        "middle": [],
                        "last": "Mehdad",
                        "suffix": ""
                    },
                    {
                        "first": "Giuseppe",
                        "middle": [],
                        "last": "Carenini",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Raymond",
                        "suffix": ""
                    },
                    {
                        "first": "Bita",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Nejat",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP)",
                "volume": "",
                "issue": "",
                "pages": "1602--1613",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Shima Gerani, Yashar Mehdad, Giuseppe Carenini, Raymond T Ng, and Bita Nejat. 2014. Abstractive summarization of product reviews using discourse structure. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1602-1613.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Gaussian error linear units (gelus)",
                "authors": [
                    {
                        "first": "Dan",
                        "middle": [],
                        "last": "Hendrycks",
                        "suffix": ""
                    },
                    {
                        "first": "Kevin",
                        "middle": [],
                        "last": "Gimpel",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "arXiv": [
                        "arXiv:1606.08415"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Dan Hendrycks and Kevin Gimpel. 2016. Gaus- sian error linear units (gelus). arXiv preprint arXiv:1606.08415.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "Hilda: A discourse parser using support vector machine classification",
                "authors": [
                    {
                        "first": "Hugo",
                        "middle": [],
                        "last": "Hernault",
                        "suffix": ""
                    },
                    {
                        "first": "Helmut",
                        "middle": [],
                        "last": "Prendinger",
                        "suffix": ""
                    },
                    {
                        "first": "Mitsuru",
                        "middle": [],
                        "last": "Ishizuka",
                        "suffix": ""
                    }
                ],
                "year": 2010,
                "venue": "Dialogue & Discourse",
                "volume": "1",
                "issue": "3",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hugo Hernault, Helmut Prendinger, Mitsuru Ishizuka, et al. 2010. Hilda: A discourse parser using sup- port vector machine classification. Dialogue & Dis- course, 1(3).",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Using rhetorical structure in sentiment analysis",
                "authors": [
                    {
                        "first": "Alexander",
                        "middle": [],
                        "last": "Hogenboom",
                        "suffix": ""
                    },
                    {
                        "first": "Flavius",
                        "middle": [],
                        "last": "Frasincar",
                        "suffix": ""
                    },
                    {
                        "first": "Franciska",
                        "middle": [
                            "De"
                        ],
                        "last": "Jong",
                        "suffix": ""
                    },
                    {
                        "first": "Uzay",
                        "middle": [],
                        "last": "Kaymak",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Commun. ACM",
                "volume": "58",
                "issue": "7",
                "pages": "69--77",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Alexander Hogenboom, Flavius Frasincar, Franciska De Jong, and Uzay Kaymak. 2015. Using rhetori- cal structure in sentiment analysis. Commun. ACM, 58(7):69-77.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Discourse-sentiment alignment tool (dsat)",
                "authors": [
                    {
                        "first": "Patrick",
                        "middle": [],
                        "last": "Huber",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Patrick Huber. 2019. Discourse-sentiment alignment tool (dsat).",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Representation learning for text-level discourse parsing",
                "authors": [
                    {
                        "first": "Yangfeng",
                        "middle": [],
                        "last": "Ji",
                        "suffix": ""
                    },
                    {
                        "first": "Jacob",
                        "middle": [],
                        "last": "Eisenstein",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "13--24",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yangfeng Ji and Jacob Eisenstein. 2014. Representa- tion learning for text-level discourse parsing. In Pro- ceedings of the 52nd Annual Meeting of the Associa- tion for Computational Linguistics (Volume 1: Long Papers), volume 1, pages 13-24.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Neural discourse structure for text categorization",
                "authors": [
                    {
                        "first": "Yangfeng",
                        "middle": [],
                        "last": "Ji",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Noah",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Smith",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "996--1005",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Yangfeng Ji and Noah A Smith. 2017. Neural dis- course structure for text categorization. In Proceed- ings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pa- pers), pages 996-1005.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Spanbert: Improving pre-training by representing and predicting spans",
                "authors": [
                    {
                        "first": "Mandar",
                        "middle": [],
                        "last": "Joshi",
                        "suffix": ""
                    },
                    {
                        "first": "Danqi",
                        "middle": [],
                        "last": "Chen",
                        "suffix": ""
                    },
                    {
                        "first": "Yinhan",
                        "middle": [],
                        "last": "Liu",
                        "suffix": ""
                    },
                    {
                        "first": "Daniel",
                        "middle": [
                            "S"
                        ],
                        "last": "Weld",
                        "suffix": ""
                    },
                    {
                        "first": "Luke",
                        "middle": [],
                        "last": "Zettlemoyer",
                        "suffix": ""
                    },
                    {
                        "first": "Omer",
                        "middle": [],
                        "last": "Levy",
                        "suffix": ""
                    }
                ],
                "year": 2020,
                "venue": "Transactions of the Association for Computational Linguistics",
                "volume": "8",
                "issue": "",
                "pages": "64--77",
                "other_ids": {
                    "DOI": [
                        "10.1162/tacl_a_00300"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, and Omer Levy. 2020. Spanbert: Improving pre-training by representing and predicting spans. Transactions of the Associa- tion for Computational Linguistics, 8:64-77.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Codra: A novel discriminative framework for rhetorical analysis",
                "authors": [
                    {
                        "first": "Shafiq",
                        "middle": [],
                        "last": "Joty",
                        "suffix": ""
                    },
                    {
                        "first": "Giuseppe",
                        "middle": [],
                        "last": "Carenini",
                        "suffix": ""
                    },
                    {
                        "first": "Raymond T",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Computational Linguistics",
                "volume": "",
                "issue": "3",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Shafiq Joty, Giuseppe Carenini, and Raymond T Ng. 2015. Codra: A novel discriminative framework for rhetorical analysis. Computational Linguistics, 41(3).",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Top-down rst parsing utilizing granularity levels in documents",
                "authors": [
                    {
                        "first": "Naoki",
                        "middle": [],
                        "last": "Kobayashi",
                        "suffix": ""
                    },
                    {
                        "first": "Tsutomu",
                        "middle": [],
                        "last": "Hirao",
                        "suffix": ""
                    },
                    {
                        "first": "Hidetaka",
                        "middle": [],
                        "last": "Kamigaito",
                        "suffix": ""
                    },
                    {
                        "first": "Manabu",
                        "middle": [],
                        "last": "Okumura",
                        "suffix": ""
                    },
                    {
                        "first": "Masaaki",
                        "middle": [],
                        "last": "Nagata",
                        "suffix": ""
                    }
                ],
                "year": 2020,
                "venue": "Proceedings of the AAAI Conference on Artificial Intelligence",
                "volume": "34",
                "issue": "",
                "pages": "8099--8106",
                "other_ids": {
                    "DOI": [
                        "10.1609/aaai.v34i05.6321"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Naoki Kobayashi, Tsutomu Hirao, Hidetaka Kami- gaito, Manabu Okumura, and Masaaki Nagata. 2020. Top-down rst parsing utilizing granularity levels in documents. Proceedings of the AAAI Conference on Artificial Intelligence, 34:8099-8106.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "Higher-order coreference resolution with coarse-tofine inference",
                "authors": [
                    {
                        "first": "Kenton",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "Luheng",
                        "middle": [],
                        "last": "He",
                        "suffix": ""
                    },
                    {
                        "first": "Luke",
                        "middle": [],
                        "last": "Zettlemoyer",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
                "volume": "2",
                "issue": "",
                "pages": "687--692",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/N18-2108"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Kenton Lee, Luheng He, and Luke Zettlemoyer. 2018. Higher-order coreference resolution with coarse-to- fine inference. In Proceedings of the 2018 Confer- ence of the North American Chapter of the Associ- ation for Computational Linguistics: Human Lan- guage Technologies, Volume 2 (Short Papers), pages 687-692, New Orleans, Louisiana. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Decoupled weight decay regularization",
                "authors": [
                    {
                        "first": "Ilya",
                        "middle": [],
                        "last": "Loshchilov",
                        "suffix": ""
                    },
                    {
                        "first": "Frank",
                        "middle": [],
                        "last": "Hutter",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "International Conference on Learning Representations",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. In International Con- ference on Learning Representations.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Rhetorical structure theory: Toward a functional theory of text organization. Text-Interdisciplinary Journal for the Study of Discourse",
                "authors": [
                    {
                        "first": "C",
                        "middle": [],
                        "last": "William",
                        "suffix": ""
                    },
                    {
                        "first": "Sandra",
                        "middle": [
                            "A"
                        ],
                        "last": "Mann",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Thompson",
                        "suffix": ""
                    }
                ],
                "year": 1988,
                "venue": "",
                "volume": "8",
                "issue": "",
                "pages": "243--281",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "William C Mann and Sandra A Thompson. 1988. Rhetorical structure theory: Toward a functional the- ory of text organization. Text-Interdisciplinary Jour- nal for the Study of Discourse, 8(3):243-281.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "The Theory and Practice of Discourse Parsing and Summarization",
                "authors": [
                    {
                        "first": "Daniel",
                        "middle": [],
                        "last": "Marcu",
                        "suffix": ""
                    }
                ],
                "year": 2000,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "DOI": [
                        "10.7551/mitpress/6754.001.0001"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Daniel Marcu. 2000. The Theory and Practice of Dis- course Parsing and Summarization.",
                "links": null
            },
            "BIBREF24": {
                "ref_id": "b24",
                "title": "How much progress have we made on RST discourse parsing? a replication study of recent results on the RST-DT",
                "authors": [
                    {
                        "first": "Mathieu",
                        "middle": [],
                        "last": "Morey",
                        "suffix": ""
                    },
                    {
                        "first": "Philippe",
                        "middle": [],
                        "last": "Muller",
                        "suffix": ""
                    },
                    {
                        "first": "Nicholas",
                        "middle": [],
                        "last": "Asher",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "1319--1324",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/D17-1136"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Mathieu Morey, Philippe Muller, and Nicholas Asher. 2017. How much progress have we made on RST discourse parsing? a replication study of recent re- sults on the RST-DT. In Proceedings of the 2017 Conference on Empirical Methods in Natural Lan- guage Processing, pages 1319-1324, Copenhagen, Denmark. Association for Computational Linguis- tics.",
                "links": null
            },
            "BIBREF25": {
                "ref_id": "b25",
                "title": "A dependency perspective on rst discourse parsing and evaluation",
                "authors": [
                    {
                        "first": "Mathieu",
                        "middle": [],
                        "last": "Morey",
                        "suffix": ""
                    },
                    {
                        "first": "Philippe",
                        "middle": [],
                        "last": "Muller",
                        "suffix": ""
                    },
                    {
                        "first": "Nicholas",
                        "middle": [],
                        "last": "Asher",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Computational Linguistics",
                "volume": "44",
                "issue": "2",
                "pages": "197--235",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Mathieu Morey, Philippe Muller, and Nicholas Asher. 2018. A dependency perspective on rst discourse parsing and evaluation. Computational Linguistics, 44(2):197-235.",
                "links": null
            },
            "BIBREF26": {
                "ref_id": "b26",
                "title": "Exploring joint neural model for sentence level discourse parsing and sentiment analysis",
                "authors": [
                    {
                        "first": "Bita",
                        "middle": [],
                        "last": "Nejat",
                        "suffix": ""
                    },
                    {
                        "first": "Giuseppe",
                        "middle": [],
                        "last": "Carenini",
                        "suffix": ""
                    },
                    {
                        "first": "Raymond",
                        "middle": [],
                        "last": "Ng",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue",
                "volume": "",
                "issue": "",
                "pages": "289--298",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Bita Nejat, Giuseppe Carenini, and Raymond Ng. 2017. Exploring joint neural model for sentence level dis- course parsing and sentiment analysis. In Proceed- ings of the 18th Annual SIGdial Meeting on Dis- course and Dialogue, pages 289-298.",
                "links": null
            },
            "BIBREF27": {
                "ref_id": "b27",
                "title": "Pytorch: An imperative style, high-performance deep learning library",
                "authors": [
                    {
                        "first": "Adam",
                        "middle": [],
                        "last": "Paszke",
                        "suffix": ""
                    },
                    {
                        "first": "Sam",
                        "middle": [],
                        "last": "Gross",
                        "suffix": ""
                    },
                    {
                        "first": "Francisco",
                        "middle": [],
                        "last": "Massa",
                        "suffix": ""
                    },
                    {
                        "first": "Adam",
                        "middle": [],
                        "last": "Lerer",
                        "suffix": ""
                    },
                    {
                        "first": "James",
                        "middle": [],
                        "last": "Bradbury",
                        "suffix": ""
                    },
                    {
                        "first": "Gregory",
                        "middle": [],
                        "last": "Chanan",
                        "suffix": ""
                    },
                    {
                        "first": "Trevor",
                        "middle": [],
                        "last": "Killeen",
                        "suffix": ""
                    },
                    {
                        "first": "Zeming",
                        "middle": [],
                        "last": "Lin",
                        "suffix": ""
                    },
                    {
                        "first": "Natalia",
                        "middle": [],
                        "last": "Gimelshein",
                        "suffix": ""
                    },
                    {
                        "first": "Luca",
                        "middle": [],
                        "last": "Antiga",
                        "suffix": ""
                    },
                    {
                        "first": "Alban",
                        "middle": [],
                        "last": "Desmaison",
                        "suffix": ""
                    },
                    {
                        "first": "Andreas",
                        "middle": [],
                        "last": "Kopf",
                        "suffix": ""
                    },
                    {
                        "first": "Edward",
                        "middle": [],
                        "last": "Yang",
                        "suffix": ""
                    },
                    {
                        "first": "Zachary",
                        "middle": [],
                        "last": "Devito",
                        "suffix": ""
                    },
                    {
                        "first": "Martin",
                        "middle": [],
                        "last": "Raison",
                        "suffix": ""
                    },
                    {
                        "first": "Alykhan",
                        "middle": [],
                        "last": "Tejani",
                        "suffix": ""
                    },
                    {
                        "first": "Sasank",
                        "middle": [],
                        "last": "Chilamkurthy",
                        "suffix": ""
                    },
                    {
                        "first": "Benoit",
                        "middle": [],
                        "last": "Steiner",
                        "suffix": ""
                    },
                    {
                        "first": "Lu",
                        "middle": [],
                        "last": "Fang",
                        "suffix": ""
                    },
                    {
                        "first": "Junjie",
                        "middle": [],
                        "last": "Bai",
                        "suffix": ""
                    },
                    {
                        "first": "Soumith",
                        "middle": [],
                        "last": "Chintala",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "Advances in Neural Information Processing Systems",
                "volume": "32",
                "issue": "",
                "pages": "8026--8037",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Te- jani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. 2019. Py- torch: An imperative style, high-performance deep learning library. In H. Wallach, H. Larochelle, A. Beygelzimer, F. dAlch\u00e9-Buc, E. Fox, and R. Gar- nett, editors, Advances in Neural Information Pro- cessing Systems 32, pages 8026-8037. Curran Asso- ciates, Inc.",
                "links": null
            },
            "BIBREF28": {
                "ref_id": "b28",
                "title": "Deep contextualized word representations",
                "authors": [
                    {
                        "first": "Matthew",
                        "middle": [],
                        "last": "Peters",
                        "suffix": ""
                    },
                    {
                        "first": "Mark",
                        "middle": [],
                        "last": "Neumann",
                        "suffix": ""
                    },
                    {
                        "first": "Mohit",
                        "middle": [],
                        "last": "Iyyer",
                        "suffix": ""
                    },
                    {
                        "first": "Matt",
                        "middle": [],
                        "last": "Gardner",
                        "suffix": ""
                    },
                    {
                        "first": "Christopher",
                        "middle": [],
                        "last": "Clark",
                        "suffix": ""
                    },
                    {
                        "first": "Kenton",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    },
                    {
                        "first": "Luke",
                        "middle": [],
                        "last": "Zettlemoyer",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
                "volume": "1",
                "issue": "",
                "pages": "2227--2237",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/N18-1202"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word rep- resentations. In Proceedings of the 2018 Confer- ence of the North American Chapter of the Associ- ation for Computational Linguistics: Human Lan- guage Technologies, Volume 1 (Long Papers), pages 2227-2237, New Orleans, Louisiana. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF29": {
                "ref_id": "b29",
                "title": "A hierarchical multi-task approach for learning embeddings from semantic tasks",
                "authors": [
                    {
                        "first": "Victor",
                        "middle": [],
                        "last": "Sanh",
                        "suffix": ""
                    },
                    {
                        "first": "Thomas",
                        "middle": [],
                        "last": "Wolf",
                        "suffix": ""
                    },
                    {
                        "first": "Sebastian",
                        "middle": [],
                        "last": "Ruder",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Victor Sanh, Thomas Wolf, and Sebastian Ruder. 2018. A hierarchical multi-task approach for learning em- beddings from semantic tasks.",
                "links": null
            },
            "BIBREF30": {
                "ref_id": "b30",
                "title": "A two-stage parsing method for text-level discourse analysis",
                "authors": [
                    {
                        "first": "Yizhong",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    },
                    {
                        "first": "Sujian",
                        "middle": [],
                        "last": "Li",
                        "suffix": ""
                    },
                    {
                        "first": "Houfeng",
                        "middle": [],
                        "last": "Wang",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics",
                "volume": "2",
                "issue": "",
                "pages": "184--188",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/P17-2029"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yizhong Wang, Sujian Li, and Houfeng Wang. 2017. A two-stage parsing method for text-level discourse analysis. In Proceedings of the 55th Annual Meet- ing of the Association for Computational Linguistics (Volume 2: Short Papers), pages 184-188, Vancou- ver, Canada. Association for Computational Linguis- tics.",
                "links": null
            },
            "BIBREF32": {
                "ref_id": "b32",
                "title": "Transition-based neural rst parsing with implicit syntax features",
                "authors": [
                    {
                        "first": "Nan",
                        "middle": [],
                        "last": "Yu",
                        "suffix": ""
                    },
                    {
                        "first": "Meishan",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Guohong",
                        "middle": [],
                        "last": "Fu",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 27th International Conference on Computational Linguistics",
                "volume": "",
                "issue": "",
                "pages": "559--570",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Nan Yu, Meishan Zhang, and Guohong Fu. 2018. Transition-based neural rst parsing with implicit syn- tax features. In Proceedings of the 27th Inter- national Conference on Computational Linguistics, pages 559-570.",
                "links": null
            },
            "BIBREF33": {
                "ref_id": "b33",
                "title": "A distributional view of discourse encapsulation: Multifactorial prediction of coreference density in RST",
                "authors": [
                    {
                        "first": "Amir",
                        "middle": [],
                        "last": "Zeldes",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 6th Workshop on Recent Advances in RST and Related Formalisms",
                "volume": "",
                "issue": "",
                "pages": "20--28",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/W17-3603"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Amir Zeldes. 2017. A distributional view of discourse encapsulation: Multifactorial prediction of corefer- ence density in RST. In Proceedings of the 6th Workshop on Recent Advances in RST and Related Formalisms, pages 20-28, Santiago de Compostela, Spain. Association for Computational Linguistics.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "type_str": "figure",
                "text": "An example",
                "num": null,
                "uris": null
            },
            "FIGREF1": {
                "type_str": "figure",
                "text": "Overview of our models. For spans S 2 are S 1 , the w i:i+2 and w j,j+1 respectively are the nuclear EDUs. (Left) All components of SpanBERT-NoCoref. (Middle) SpanBERT-CorefFeats modifies the initial SpanBERT embeddings according to predicted coreference clusters (in red). (Right) SpanBERT-MultiTask updates the final span representations with embeddings of mentions of shared entities.",
                "num": null,
                "uris": null
            },
            "FIGREF2": {
                "type_str": "figure",
                "text": "A subtree from SpanBERT-NoCoref prediction for wsj 0631 (left) and gold-standard (right",
                "num": null,
                "uris": null
            },
            "TABREF0": {
                "html": null,
                "text": "The leaf EDUs are as follows: [Max had a great evening last night.] 1 [He had a great meal.] 2 [He ate salmon.] 3 [He devoured lots of cheese.] 4 [He then won a dancing competition.] 5 [It was a beautiful pink] 6",
                "content": "<table/>",
                "type_str": "table",
                "num": null
            },
            "TABREF1": {
                "html": null,
                "text": "Parseval micro precision for structure, nuclearity and relation prediction on RST-DT corpus. Scores for previous approaches are from eitherMorey et al. (2017) or the original papers.",
                "content": "<table><tr><td>Model</td><td colspan=\"2\">Structure Nuclearity</td><td>Relation</td></tr><tr><td>HILDA(2010)</td><td>82.6</td><td>66.6</td><td>54.6</td></tr><tr><td>DPLP(2014)</td><td>82.0</td><td>68.2</td><td>57.8</td></tr><tr><td>CODRA(2015)</td><td>82.6</td><td>68.3</td><td>55.8</td></tr><tr><td>Two-Stage(2017)</td><td>86.0</td><td>72.4</td><td>59.7</td></tr><tr><td>Transition-Syntax(2018)</td><td>85.5</td><td>73.1</td><td>60.2</td></tr><tr><td>D2P2S2E (Ensemble)(2020)</td><td>87.0</td><td>74.6</td><td>60.0</td></tr><tr><td>SpanBERT-NoCoref</td><td colspan=\"3\">87.8 \u00b1 0.2 75.8 \u00b1 0.2 63.4 \u00b1 0.3</td></tr><tr><td>SpanBERT-CorefFeats</td><td colspan=\"3\">88.1 \u00b1 0.3 76.1 \u00b1 0.6 63.6 \u00b1 0.3</td></tr><tr><td>SpanBERT-MultiTask</td><td colspan=\"3\">87.9 \u00b1 0.2 75.9 \u00b1 0.6 63.3 \u00b1 0.7</td></tr><tr><td>Human (2017)</td><td>88.3</td><td>77.3</td><td>65.4</td></tr><tr><td>Table 1: RST-</td><td/><td/><td/></tr></table>",
                "type_str": "table",
                "num": null
            }
        }
    }
}