File size: 90,797 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 |
{
"paper_id": "I13-1025",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T07:14:49.823119Z"
},
"title": "Written Dialog and Social Power: Manifestations of Different Types of Power in Dialog Behavior",
"authors": [
{
"first": "Vinodkumar",
"middle": [],
"last": "Prabhakaran",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Columbia University",
"location": {
"settlement": "New York",
"region": "NY"
}
},
"email": ""
},
{
"first": "Owen",
"middle": [],
"last": "Rambow",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Columbia University",
"location": {
"settlement": "New York",
"region": "NY"
}
},
"email": "rambow@ccls.columbia.edu"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Dialog behavior is affected by power relations among the discourse participants. We show that four different types of power relations (hierarchical power, situational power, influence, and power over communication) affect written dialog behavior in different ways. We also present a system that can identify power relations given a written dialog.",
"pdf_parse": {
"paper_id": "I13-1025",
"_pdf_hash": "",
"abstract": [
{
"text": "Dialog behavior is affected by power relations among the discourse participants. We show that four different types of power relations (hierarchical power, situational power, influence, and power over communication) affect written dialog behavior in different ways. We also present a system that can identify power relations given a written dialog.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The recent increase in online social interactions has triggered great interest in computationally analyzing such interactions to gain insights about the discourse participants (DPs). Within the field of analyzing online interactions, there is a growing interest in finding how social power relations between participants are reflected in the various facets of interactions, and whether the power relations can be detected using computational means (Rowe et al., 2007; Bramsen et al., 2011) . More recent work has shown that an analysis of the dialog structure (and not just the message content) helps detecting power relations (Biran et al., 2012; Danescu-Niculescu-Mizil et al., 2012) .",
"cite_spans": [
{
"start": 448,
"end": 467,
"text": "(Rowe et al., 2007;",
"ref_id": "BIBREF21"
},
{
"start": 468,
"end": 489,
"text": "Bramsen et al., 2011)",
"ref_id": "BIBREF3"
},
{
"start": 627,
"end": 647,
"text": "(Biran et al., 2012;",
"ref_id": "BIBREF1"
},
{
"start": 648,
"end": 685,
"text": "Danescu-Niculescu-Mizil et al., 2012)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Understanding the relation between dialog and power may help in various applications. For example, if a dialog system is engineered to behave appropriately given the user's expectation of relative power (for different types of power), then the user may experience the interaction with the system as more natural. Turning to dialog analysis rather than generation, we can build a computational system to analyze power relations between participants in an interaction. Such a system could have various applications. Power analysis in online forums and communities could be useful in determining relevance to a user searching the forum. For example, a user may want to limit his search to posts authored by the DPs with higher power. Power analysis may also aid law enforcement agencies to detect leaders and influencers in suspicious online communities. This is especially useful since the real identities of the members of such communities are often not revealed and their hierarchies may not be available to the law enforcement agencies.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The power differential between the DPs may be based on a multitude of factors such as status, authority, role, knowledge and so on. Early computational approaches to analyzing power in interactions relied solely on static power structures such as corporate hierarchies as the source of the power differential (Rowe et al., 2007; Bramsen et al., 2011) . More recent studies have looked into dynamic notions of power as well, such as influence (Biran et al., 2012) . However, not much work has been done to understand how different types of power differ in the ways they affect how people interact in dialog.",
"cite_spans": [
{
"start": 309,
"end": 328,
"text": "(Rowe et al., 2007;",
"ref_id": "BIBREF21"
},
{
"start": 329,
"end": 350,
"text": "Bramsen et al., 2011)",
"ref_id": "BIBREF3"
},
{
"start": 442,
"end": 462,
"text": "(Biran et al., 2012)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this paper, we study four different types of power -hierarchical power, situational power, influence and power over communication. We investigate whether all four social power relations are manifested in dialog behavior; we restrict our attention to written dialog, specifically email exchanged in an American corporation. By \"dialog behavior\", we mean the choices a DP makes while engaging in dialog. Dialog behavior includes choices that affect dialog structure, such as the choice of when to participate (e.g., does the DP initiate the dialog?), how much to contribute (e.g., is the DP terse or loquacious?), what sort of contribution to make (e.g., which dialog acts does the DP perform? how does the contribution link to previous dialog contributions?), and what form the contribution should take (e.g., whether to make an overt display of power). The main contribution of this paper is to show that the four types of power we consider are in fact different from one another and that they affect the DPs' behavior in written dialog in different but predictable ways. We analyze these manifestations in the language as well as the dialog structure of interactions. We also present a system to detect the DPs with one of these types of power from threaded email interactions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In Section 2, we discuss related work in the field. Section 3-4 presents the data, annotations, and inter-rater agreement studies on the annotations. Section 5 summarizes the dimensions of interactions we analyze. We then present the main contributions of this paper: Section 6 analyzes the variations in the manifestations of power among the four types, and Section 7 describes a system to predict persons with any of the four types of power. We then conclude and discuss future work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Within the dialog community, researchers have studied notions of control and initiative in dialogs (e.g. (Walker and Whittaker, 1990; Jordan and Di Eugenio, 1997) ). Walker and Whittaker (1990) define \"control of communication\" in terms of whether the discourse participants are providing new, unsolicited information. They use utterance level rules to determine which discourse participant (whether the speaker or the hearer) is in control, and extend it to segments of discourse. Their notion of control differs from our notion of power over communication. They model control locally over discourse segments. What we are interested in (and what our annotations capture) is the possession of controlling power by one (or more) participant(s) across the entire dialog, i.e. how a participant controls the communication in a dialog thread in order to achieve its intended goals. Despite this difference in definition, we show in Section 6 that our notion of power over communication correlates with Walker and Whittaker (1990) 's notion of control over discourse segments. Jordan and Di Eugenio (1997) suggest that \"initiative\" applies to the level of problem solving, just as \"control\" applies to the dialog level. We leave the investigation into the relation between initiative and situational power for future work.",
"cite_spans": [
{
"start": 105,
"end": 133,
"text": "(Walker and Whittaker, 1990;",
"ref_id": "BIBREF22"
},
{
"start": 134,
"end": 162,
"text": "Jordan and Di Eugenio, 1997)",
"ref_id": "BIBREF10"
},
{
"start": 166,
"end": 193,
"text": "Walker and Whittaker (1990)",
"ref_id": "BIBREF22"
},
{
"start": 998,
"end": 1025,
"text": "Walker and Whittaker (1990)",
"ref_id": "BIBREF22"
},
{
"start": 1072,
"end": 1100,
"text": "Jordan and Di Eugenio (1997)",
"ref_id": "BIBREF10"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "In social sciences, different typologies of power have been proposed. Wartenberg (1990) makes the distinction between power-over and power-to in the context of interactions. Power-over refers to relationships between interactants set by external power structures, while power-to refers to the ability an interactant possesses within the interaction, even if it is temporary. Our notions of hierarchical power and influence are special cases of power-over. Hierarchical power is determined by organizational hierarchy, while influence is determined by knowledge, expertise etc. Similarly, our notions of situational power and power over communication are special cases of power-to. Situational power applies to the situation or task at hand, while power over communication applies to the interaction itself. French and Raven (1959) proposed five bases of power: Coercive, Reward, Positional, Referent, and Expert. They are widely used to study power in sociology. We consider hierarchical power, situational power and power over communication to be positional in nature; although the former two can also have bases in coercion and rewards. The bases of influence are mainly referent and expert power.",
"cite_spans": [
{
"start": 70,
"end": 87,
"text": "Wartenberg (1990)",
"ref_id": "BIBREF23"
},
{
"start": 807,
"end": 830,
"text": "French and Raven (1959)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Studies in sociolinguistics have also explored the relation between dialog behavior and social power. O' Barr (1982) shows that power relations are manifested in language use in courtroom dialogs. Locher (2004) studies politeness in dialogs in relation to the exercise of power. The correlation between discourse structure and perceived influence of participants has also been studied (Ng et al., 1993; Ng et al., 1995) . Specifically, factors such as frequency of contribution, proportion of turns, and number of successful interruptions have been identified as important indicators of influence (Reid and Ng, 2000) . This work was done entirely on spoken dialog. In our work, we show that the core insight -conversation is a resource for influence -carries over to written dialog; we also show that it carries over to other forms of power. However, some of the characteristics of spoken dialog do not carry over directly to written dialog, most prominently among them the issue of interruptions: there is no interruption in written dialog.",
"cite_spans": [
{
"start": 105,
"end": 116,
"text": "Barr (1982)",
"ref_id": null
},
{
"start": 197,
"end": 210,
"text": "Locher (2004)",
"ref_id": "BIBREF12"
},
{
"start": 385,
"end": 402,
"text": "(Ng et al., 1993;",
"ref_id": "BIBREF13"
},
{
"start": 403,
"end": 419,
"text": "Ng et al., 1995)",
"ref_id": "BIBREF14"
},
{
"start": 597,
"end": 616,
"text": "(Reid and Ng, 2000)",
"ref_id": "BIBREF20"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "We now look at various computational approaches to extract power relations from online dialogs. Several studies have used Social Network Analysis (e.g., (Rowe et al., 2007) ) to extract social relations from online communication.",
"cite_spans": [
{
"start": 153,
"end": 172,
"text": "(Rowe et al., 2007)",
"ref_id": "BIBREF21"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Researchers have also applied NLP techniques on message content to detect power relations. Earlier approaches used simple lexical features (e.g. (Bramsen et al., 2011; Gilbert, 2012) ) while later studies have performed deeper discourse analysis and used features such as linguistic coordination (Danescu-Niculescu-Mizil et al., 2012) , language uses such as attempts to persuade and various other dialog patterns (Biran et al., 2012) . We present a more detailed discussion of the above mentioned studies and how they differ from our line of research in (Prabhakaran et al., 2012c) .",
"cite_spans": [
{
"start": 145,
"end": 167,
"text": "(Bramsen et al., 2011;",
"ref_id": "BIBREF3"
},
{
"start": 168,
"end": 182,
"text": "Gilbert, 2012)",
"ref_id": "BIBREF6"
},
{
"start": 296,
"end": 334,
"text": "(Danescu-Niculescu-Mizil et al., 2012)",
"ref_id": "BIBREF4"
},
{
"start": 414,
"end": 434,
"text": "(Biran et al., 2012)",
"ref_id": "BIBREF1"
},
{
"start": 555,
"end": 582,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Our research also falls into the category of studies that go beyond pure lexical features and use dialog structure based features to extract social power relations. In (Prabhakaran et al., 2012c) , we studied the notion of situational power in depth and presented a system to detect persons with situational power using dialog features. In this paper as well, we use the system described in (Prabhakaran et al., 2012c) . However, this work differs from (Prabhakaran et al., 2012c) and other studies described above in that our focus is on how different types of power are manifested differently in the dialog behavior of the participants. We show that the types of power we consider are in fact different and vary in the ways they manifest in dialogs (Section 6). We also present a system that predicts different types of power (Section 7), not just hierarchical or situational power.",
"cite_spans": [
{
"start": 168,
"end": 195,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
},
{
"start": 391,
"end": 418,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
},
{
"start": 453,
"end": 480,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "We use the subset of the Enron email corpus with power annotations presented in Prabhakaran et al. (2012a) for our experiments. The corpus also contains manual dialog act annotations by Hu et al. (2009) , which enable us to perform the analysis of how power affects dialog behavior. The corpus contains 122 email threads with a total of 360 messages and 20,740 word tokens. There are about 8.5 participants per thread. There are 221 active participants (participants of a thread who has sent at least one email message in the thread) in the corpus. Table 1 presents the counts and percentages of active participants with each type of power in the corpus. We now define the four types of power we investigate in this paper. Hierarchical Power (HP): We use the gold organizational hierarchy for Enron released by Agarwal et al. (2012) to model hierarchical power. It contains relations between 1,518 employees, and Table 1 : Annotation statistics 13,724 dominance pairs (pairs of employees such that the first dominates the second in the hierarchy, not necessarily immediately). We labeled a participant to have hierarchical power within a thread if there exist a dominance pair in the gold hierarchy such that he/she dominates any other participant in the same thread. For the other three types of power -situational power, power over communication, and influence, we utilize the manual annotations present in the corpus of (Prabhakaran et al., 2012a) . 1 We labeled a participant to have one of these types of power within a thread if he or she was judged to have that type of power over any other participant in the same thread. We explain the annotations in detail below with an example thread and corresponding annotations shown in Table 2 ; the email body contains dialog act and link annotations in [square brackets] which will be explained in Section 3.1. Situational Power (SP): Person 1 is said to have situational power over person 2 if person 1 has power or authority to direct and/or approve person 2 's actions in the current situation or while a particular task is being performed, based on the communication in the current thread. Situational power is independent of organizational hierarchy: person 1 with situational power may or may not be above person 2 in the organizational hierarchy (or there may be no organizational hierarchy at all). In our example thread, our annotator judged Kathryn to posses situational power over Leslie, Sara and Brent because Kathryn is following up on and assigning a task to others, and because Kathryn uses language that shows that she is in charge of the situation. Table 2 : Example thread with power annotations people who simply respond to questions or perform actions when directed to do so. There could be multiple such participants in a given thread. In our example thread, no one was judged to have power over communication since the communication is broken into two separate interactions of just one message each -one from Kathryn to everyone and the other between Brent and Mark.",
"cite_spans": [
{
"start": 80,
"end": 106,
"text": "Prabhakaran et al. (2012a)",
"ref_id": "BIBREF17"
},
{
"start": 186,
"end": 202,
"text": "Hu et al. (2009)",
"ref_id": "BIBREF7"
},
{
"start": 811,
"end": 832,
"text": "Agarwal et al. (2012)",
"ref_id": "BIBREF0"
},
{
"start": 1423,
"end": 1450,
"text": "(Prabhakaran et al., 2012a)",
"ref_id": "BIBREF17"
},
{
"start": 1453,
"end": 1454,
"text": "1",
"ref_id": null
}
],
"ref_spans": [
{
"start": 549,
"end": 556,
"text": "Table 1",
"ref_id": null
},
{
"start": 913,
"end": 920,
"text": "Table 1",
"ref_id": null
},
{
"start": 1735,
"end": 1742,
"text": "Table 2",
"ref_id": null
},
{
"start": 2618,
"end": 2625,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Data and Annotations",
"sec_num": "3"
},
{
"text": "A person is defined to have influence if she 1) has credibility in the group, 2) persists in attempting to convince others, even if some disagreement occurs, 3) introduces topics/ideas that others pick up on or support, and 4) is a group participant but not necessarily active in the discussion(s) where others support/credit her. In addition, the influencer's ideas or language may be adopted by others and others may explicitly recognize influencer's authority. In our example, our annotator judged Mark to have influence over Brent since the latter seeks advice from the former on how to deal with the situation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Influence (INFL):",
"sec_num": null
},
{
"text": "The corpus we used contains manual dialog act annotations as described in Hu et al. (2009) . We use these annotations to model the dialog structure of the communication thread. For each message, Hu et al. (2009) assign a Dialog Act (DA) label to each segment of text with a coherent communicative function. The label could be one of the following: ReqAction, ReqInfo, Inform, In-formOffline, 3 Conventional, and Commit. In addition, the segments are linked by three types of links to reflect the dialog structure. These links capture the patterns of local alternation between an initiating dialog act and a responding one. A forward link (Flink) is the analog of a \"first pairpart\" of an adjacency pair, is restricted to ReqInfo and ReqAction segments. The responses to such requests are assigned a backward link (Blink). If an utterance can be interpreted as a response to a preceding segment, it gets a Blink even where the preceding segment has no Flink. The preceding segment taken to be the \"first pair-part\" of the link is assigned a secondary forward link (SFlink).",
"cite_spans": [
{
"start": 74,
"end": 90,
"text": "Hu et al. (2009)",
"ref_id": "BIBREF7"
},
{
"start": 195,
"end": 211,
"text": "Hu et al. (2009)",
"ref_id": "BIBREF7"
},
{
"start": 392,
"end": 393,
"text": "3",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Dialog Act Annotations",
"sec_num": "3.1"
},
{
"text": "Our corpus also contains the overt display of power (ODP) (Prabhakaran et al., 2012b) annotations. An utterance is defined to have an ODP if it is interpreted as creating additional constraints on the response beyond those imposed by the general dialog act. Syntactically, an ODP can be an imperative, a question, or a declarative sentence. In our example thread, utterance M1.2 is an instance of ODP. The inter-annotator agreement value (\u03ba) of ODP annotations was 0.67.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Overt Display of Power",
"sec_num": "3.2"
},
{
"text": "The power annotations in the corpus are performed by a single annotator and capture her perception of the overall power structure among the participants of the interaction. To verify the reliability of these annotations, we performed an independent inter-annotator agreement (IAA) study on a subset of 47 threads from the corpus. We trained two annotators -AnnA and AnnB -using the same annotation manual described in (Prabhakaran et al., 2012a ) and compared the annotations they produced on the selected threads. Annotators were asked to read the entire thread before performing the annotations. They are also asked to provide, in free-form English, a short \"power narrative\" which describes their perception of the overall power structure among the discourse participants of that thread. Annotators build a fairly consistent mental image of a power narrative -an outline of the power structure between the participants -based on various indicators from across the thread. Their individual power annotations are based on this power narrative. Hence, the cognitive process behind labeling a participant to have a particular type of power is not a binary decision the annotator makes for each participant. However, evaluating agreement on such a formulation is not straightforward. Thus, for the purpose of this IAA study, we port this task into a binary decision task of identifying whether participant X has power of type P or not.",
"cite_spans": [
{
"start": 418,
"end": 444,
"text": "(Prabhakaran et al., 2012a",
"ref_id": "BIBREF17"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Reliability of Annotations",
"sec_num": "4"
},
{
"text": "There were 289 participants in the selected 47 threads. The \u03ba values obtained for each type of power is shown in Table 3 under Round 1. Since the \u03ba values obtained in round 1 were only fair to moderate, we performed another round of training and inter annotator study. For this round, AnnB was not available, and we hired another annotator AnnC. The \u03ba values obtained between AnnA and AnnC on another set of 10 threads is presented in The \u03ba values obtained in both round 1 and round 2 are in the range of those previously reported for similar tasks (e.g., 0.18 for managerial influence and 0.52 for establishing solidarity (Bracewell et al., 2012) ; 0.72 for influence (Biran et al., 2012) ). The agreement in round 2 improved considerably for both PC and INFL after the second round of training. The issue of moderate agreement for SP and its possible reasons are discussed in detail in (Prabhakaran et al., 2012c) . For the rest of this paper, we use the original annotations that were present in the corpus.",
"cite_spans": [
{
"start": 623,
"end": 647,
"text": "(Bracewell et al., 2012)",
"ref_id": "BIBREF2"
},
{
"start": 669,
"end": 689,
"text": "(Biran et al., 2012)",
"ref_id": "BIBREF1"
},
{
"start": 888,
"end": 915,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [
{
"start": 113,
"end": 120,
"text": "Table 3",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Reliability of Annotations",
"sec_num": "4"
},
{
"text": "We use five sets of features to capture the dialog behavior of participants: dialog act percentages (DAP), dialog link counts (DLC), positional (PST), verbosity (VRB), and overt displays of power (ODP). The specific features within each set are listed in Table 4 . PST and VRB are readily derivable from the data, without any annotations.",
"cite_spans": [],
"ref_spans": [
{
"start": 255,
"end": 262,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Dialog Behavior",
"sec_num": "5"
},
{
"text": "Features DAP ReqAction, ReqInform, Inform, InformOffline, Conventional, Commit DLC Flink, SFlink, Blink, Clink, Dlink, DlinkRatio PST Initiator, FirstMsg, LastMsg VRB MsgCount, MsgRatio, TokenCount, TokenRatio, TokensPerMsg ODP ODPCount Table 4 : Feature Sets DAP captures the percentages of each dialog act labels in each participant's utterances. DLC captures the metrics on various kinds of links in each participant's messages. Flink, SFlink and Blink corresponds to counts of respective link annotations in participants' messages. We refer to Flinks with one or more backward links as connected links (Clink) and those with no matching Blink as dangling links (Dlink). A dangling link denotes a request that was ignored. The DlinkRatio is the ratio of Dlinks to Flinks for a participant. This captures what percentage of a participant's requests went unanswered. PST captures the positions within the thread where the participant joined and left the conversation. Initiator is a binary feature capturing whether the participant initiated the thread or not. FirstMsg and LastMsg are real valued features between 0 and 1, capturing the relative position of the first and last messages by the participant. VRB features are self explanatory. ODP captures the number of instances of ODP in the messages by each participant.",
"cite_spans": [],
"ref_spans": [
{
"start": 237,
"end": 244,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Set",
"sec_num": null
},
{
"text": "In this section, we present the results of a statistical analysis of the dialog features with respect to people with the four types of power. For each type of power (HP, SP, INFL and PC), we consider two populations of people who participated in the dialog: P, those judged to have that type of power, and N , those not judged to have that power. Then, for each feature, we perform a two-sample, twotailed t-test comparing means of feature values of Table 5 presents means of each feature value for both populations P and N (as \"mean(P)| mean(N )\") along with the p-value associated with the t-test as the subscript. For p < 0.05, we reject the null hypothesis and consider the feature to be statistically significant (boldfaced in Table 5 ).",
"cite_spans": [],
"ref_spans": [
{
"start": 450,
"end": 457,
"text": "Table 5",
"ref_id": null
},
{
"start": 732,
"end": 739,
"text": "Table 5",
"ref_id": null
}
],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "We find many features which are statistically significant, which suggests that power types are reflected in the dialog structure. The t-test results also show that significance of features differ considerably from one type of power to another, which suggests that different power types are reflected differently in the dialog structure, and that they are thus indeed different types of power.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "For HP, we find that people with HP are less active in threads than those without. For example, persons with hierarchical power tend to talk less within a thread (TokenRatio). They tend to start participating much later in the threads (FirstMsg) and do not initiate threads often (Initiator). SP and PC manifest in stark contrast from HP. Persons with SP and persons with PC both tend to talk more within a thread (TokenRatio). They also tend to be the initiators of the thread (Initiator) or start participating in the thread closer to the beginning (FirstMsg). SP and PC have many other features which are also statistically significant. For example, they send significantly more messages (MsgCount). They also have significantly more instances of overt displays of power (ODPCount) than others. It is interesting to note that ODP-Count was not a significant feature for HP. It suggests that bosses don't always display their power overtly when they interact. SP and PC also differ from one another. For example, those with SP tend to request actions (ReqAction) significantly more than those without. However, this was not significant in case of PC. Similarly, the number of back links (Blink) was not a significant feature for SP. But, people with PC tend to have significantly fewer back links (Blink) than those without.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "This finding -people with PC have fewer back links -is interesting, since it aligns PC with the characterization of control by Walker and Whittaker (1990) . According to them, control over a discourse segment is determined by whether the participant provide unsolicited information in the dialog or not. In the dialog act annotation scheme we use, solicited information (in other words, responses to requests and commands) places an obligatory Blink on the corresponding text segment. Hence, the fact that people with PC have significantly larger contributions to the dialog (VRB features), but with fewer back links, suggest that most of their contribution is unsolicited information. This is in line with Walker and Whittaker (1990) 's definition of control over discourse segments.",
"cite_spans": [
{
"start": 127,
"end": 154,
"text": "Walker and Whittaker (1990)",
"ref_id": "BIBREF22"
},
{
"start": 707,
"end": 734,
"text": "Walker and Whittaker (1990)",
"ref_id": "BIBREF22"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "Although INFL has fewer data points, we found a few significant features for INFL. People with INFL never request actions (ReqAction) as opposed to those with SP who request actions more frequently than others. Also, people with INFL tend to have significantly more inform utterances (Inform). They also have significantly fewer overt displays of power (ODPCount) than others, a stark contrast to those with SP and PC.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "The statistical measures presented in previous section are exploratory in nature, presenting tests on all combinations of features and power types. We do not draw theoretical conclusions from the specific combination of interactions that are found statistically significant. Hence, we did not apply any corrections for multiple tests in statistical significance for individual features. When we apply, the Bonferroni correction for multiple tests to adjust the p-value for number of test performed (threshold = 0.05|84 = 6.0E-4), 10 features would still remain statistically significant. Hence the global null hypothesis that the features we considered do not interact with the power types would still be rejected.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Variations in Manifestations of Power",
"sec_num": "6"
},
{
"text": "In this section, we present a system to predict whether a person has a given type of power in the context of an email thread. We show that different sets of features are helpful to detect different types of power. We build a separate binary classifier for each power type predicting whether or not a given participant in a communication thread has that type of power or not. Since our dataset is skewed especially for HP & INFL (with very few persons with power), we balanced our dataset by up-sampling minority class instances in the training step. This has proven useful in cases of unbalanced datasets (Japkowicz, 2000) . All results presented below have been obtained after balancing the training folds in cross validation; the test folds remain unchanged. We used the tokenizer, POS tagger, lemmatizer and SVMLight (Joachims, 1999) wrapper in the ClearTK (Ogren et al., 2008) package. The ClearTK wrapper for SVMLight internally shifts the prediction threshold based on a posterior probabilistic score calcu- We first find the best performing subset of features for each feature set by exhaustive search within the set. Once we have the best subset of each feature set, we do another round of exhaustive search combining best performers of each set to find the overall best performing feature subset. We report micro-averaged (P)recision, (R)ecall and (F)-measure on 5-fold cross validation for each power type. We experimented with a linear kernel and a quadratic kernel; the latter performed better. All results presented in this paper are obtained using a quadratic kernel. Table 6 shows cross validation results for all four types of power for each set of features. 4 The corpus was split into folds at the thread level. We present two simple baseline measures -Random and AlwaysTrue and a langauge-based baseline, LEX. In the Random baseline, we predict an active participant to have the particular type of power at random. In AlwaysTrue baseline, we always predict an active participant to have power. For LEX, we use only lexical features (unigrams and bigrams) from messages sent by each participant to train the SVM model described above. For each power type, the table also lists (in the last row) the best performing feature subset combination and corresponding results.",
"cite_spans": [
{
"start": 605,
"end": 622,
"text": "(Japkowicz, 2000)",
"ref_id": "BIBREF8"
},
{
"start": 820,
"end": 836,
"text": "(Joachims, 1999)",
"ref_id": "BIBREF9"
},
{
"start": 860,
"end": 880,
"text": "(Ogren et al., 2008)",
"ref_id": "BIBREF16"
},
{
"start": 1675,
"end": 1676,
"text": "4",
"ref_id": null
}
],
"ref_spans": [
{
"start": 1582,
"end": 1589,
"text": "Table 6",
"ref_id": "TABREF7"
}
],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "HP is hard to predict, which could partly be due to the very small number of positive training examples in the corpus. For the LEX baseline using purely word ngrams, the system did not get any correct predictions. All feature subsets outperformed the other baselines of 11.3% and 15.0% (for Random and AlwaysTrue respectively), and a combination of VRB, PST and ODP gave the best model obtaining an F measure of 29.5%.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "For SP, the best performing individual feature sets are ODP and DLC, both at or near 60.0%. While ODP gave a high precision (71.2%) model, DLC gave a high recall (75.3%) model, the combination of both gave the best performing system with an F measure of 64.4%.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "For PC, the best single feature was FirstMsg (relative position of first message). This is because the person with the power over communication is almost always the initiator of the thread. Note that the notion of PC is not defined in terms of positional features: annotators were asked to find the participants who \"actively attempt to achieve the intended goals of the communication\". It is our finding that those who are in PC were also the ones who did initiate the thread. It is also worth noting that ODP is the worst performer for PC which is in contrast with the case of SP, supporting the claim that these two types of power are in fact different.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "INFL is another very hard class to predict, again, possibly partly due to the very small number of positive training examples. The simple baseline F measures were both 9.5, while the LEX did not produce any correct predictions at all. All feature sets except PST outperformed these baseline measures. The best performance was obtained 4 Results for SP were presented in (Prabhakaran et al., 2012c) . We present them here for comparison.",
"cite_spans": [
{
"start": 335,
"end": 336,
"text": "4",
"ref_id": null
},
{
"start": 370,
"end": 397,
"text": "(Prabhakaran et al., 2012c)",
"ref_id": "BIBREF19"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "by DLC with counts of Blinks, Flinks, Dlinks and SFlinks as features.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "For assessing statistical significance of F measure improvements over baseline, we used the Approximate Randomness Test (Yeh, 2000) . We found the improvements to be statistically significant for SP (p = 0.001), HP (p=0.001) and PC (p = 0.01) with a threshold for significance at p = 0.05. However, for INFL, the improvement was not statistically significant (p = 0.3). The statistical significance of SP, HP and PC would hold even after applying Bonferroni correction for multiple tests.",
"cite_spans": [
{
"start": 120,
"end": 131,
"text": "(Yeh, 2000)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Predicting Persons with Power",
"sec_num": "7"
},
{
"text": "We studied four types of power between participants of written dialog. We have shown that these types of power are manifested very differently with respect to the features we are using, which validates our claim that these are indeed different types of power. We also presented a supervised learning system to predict persons with one of the types of power in written dialog yielding encouraging results. We have shown that dialog features are very significant in predicting power relations in online written communication.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "8"
},
{
"text": "In future work, we intend to try predicting power relations between pairs of participants. It would be interesting to see how dialog features correlate with the other direction of power; that is from a submitter to an exerciser of power. We will investigate the use of additional features related to the dialog participants, such as gender. We will also investigate using a dialog act tagger, link predictor and an ODP tagger to build a fully automatic power predicting system. We would also like to extend this work to other genres of written communication like discussion forums and blogs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "8"
},
{
"text": "The manual annotations also capture the perception of hierarchical power. In this work, we use only the actual gold hierarchy (Agarwal et al., 2012) as described above.2 In(Prabhakaran et al., 2012a), power over communication was called \"control of communication\".",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Sometimes, the Inform act refers to a previous act of communication which did not happen in the email thread itself. Such cases are marked as Offline.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "This paper is based upon work supported by the Johns Hopkins Human Language Technology Center of Excellence and by the DARPA DEFT Program. The views expressed are those of the authors and do not reflect the official policy or position of the Department of Defense or the U.S. Government. We also thank several anonymous reviewers, including three SIGDIAL reviewers, for their constructive feedback.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": "9"
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "A Comprehensive Gold Standard for the Enron Organizational Hierarchy",
"authors": [
{
"first": "A",
"middle": [],
"last": "Agarwal",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Omuya",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Harnly",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 50th Annual Meeting of the ACL",
"volume": "2",
"issue": "",
"pages": "161--165",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Agarwal, A. Omuya, A. Harnly, and O. Rambow. 2012. A Comprehensive Gold Standard for the En- ron Organizational Hierarchy. In Proceedings of the 50th Annual Meeting of the ACL (Vol. 2: Short Pa- pers), pages 161-165, Jeju Island, Korea, July. ACL.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Detecting influencers in written online conversations",
"authors": [
{
"first": "O",
"middle": [],
"last": "Biran",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Andreas",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Mckeown",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the Second Workshop on Language in Social Media",
"volume": "",
"issue": "",
"pages": "37--45",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "O. Biran, S. Rosenthal, J. Andreas, K. McKeown, and O. Rambow. 2012. Detecting influencers in written online conversations. In Proceedings of the Second Workshop on Language in Social Media, pages 37- 45, Montr\u00e9al, Canada, June. ACL.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "A motif approach for identifying pursuits of power in social discourse",
"authors": [
{
"first": "D",
"middle": [
"B"
],
"last": "Bracewell",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Tomlinson",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Wang",
"suffix": ""
}
],
"year": 2012,
"venue": "ICSC",
"volume": "",
"issue": "",
"pages": "1--8",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "D. B. Bracewell, M. Tomlinson, and H. Wang. 2012. A motif approach for identifying pursuits of power in social discourse. In ICSC, pages 1-8. IEEE Com- puter Society.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Extracting social power relationships from natural language",
"authors": [
{
"first": "P",
"middle": [],
"last": "Bramsen",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Escobar-Molano",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Patel",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Alonso",
"suffix": ""
}
],
"year": 2011,
"venue": "The 49th Annual Meeting of the ACL",
"volume": "",
"issue": "",
"pages": "773--782",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "P. Bramsen, M. Escobar-Molano, A. Patel, and R. Alonso. 2011. Extracting social power relation- ships from natural language. In The 49th Annual Meeting of the ACL, pages 773-782. ACL.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Echoes of power: language effects and power differences in social interaction",
"authors": [
{
"first": "C",
"middle": [],
"last": "Danescu-Niculescu-Mizil",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Pang",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Kleinberg",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 21st international conference on WWW",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "C. Danescu-Niculescu-Mizil, L. Lee, B. Pang, and J. Kleinberg. 2012. Echoes of power: language ef- fects and power differences in social interaction. In Proceedings of the 21st international conference on WWW, New York, NY, USA. ACM.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "The Bases of Social Power",
"authors": [
{
"first": "J",
"middle": [
"R"
],
"last": "French",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Raven",
"suffix": ""
}
],
"year": 1959,
"venue": "Studies in Social Power",
"volume": "",
"issue": "",
"pages": "150--167",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. R. French and B. Raven. 1959. The Bases of So- cial Power. In Dorwin Cartwright, editor, Studies in Social Power, pages 150-167. Univ. of Mich. Press.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Phrases that signal workplace hierarchy",
"authors": [
{
"first": "E",
"middle": [],
"last": "Gilbert",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the ACM 2012 conference on Computer Supported Cooperative Work, CSCW '12",
"volume": "",
"issue": "",
"pages": "1037--1046",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "E. Gilbert. 2012. Phrases that signal workplace hier- archy. In Proceedings of the ACM 2012 conference on Computer Supported Cooperative Work, CSCW '12, pages 1037-1046, New York, NY, USA. ACM.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Contrasting the Interaction Structure of an Email and a Telephone Corpus: A Machine Learning Approach to Annotation of Dialogue Function Units",
"authors": [
{
"first": "J",
"middle": [],
"last": "Hu",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Passonneau",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the SIGDIAL 2009 Conference",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J. Hu, R. Passonneau, and O. Rambow. 2009. Con- trasting the Interaction Structure of an Email and a Telephone Corpus: A Machine Learning Approach to Annotation of Dialogue Function Units. In Pro- ceedings of the SIGDIAL 2009 Conference, London, UK, September. ACL.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Learning from imbalanced data sets: Comparison of various strategies",
"authors": [
{
"first": "N",
"middle": [],
"last": "Japkowicz",
"suffix": ""
}
],
"year": 2000,
"venue": "AAAI Workshop on Learning from Imbalanced Data Sets",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "N. Japkowicz. 2000. Learning from imbalanced data sets: Comparison of various strategies. In AAAI Workshop on Learning from Imbalanced Data Sets.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Making Large-Scale SVM Learning Practical",
"authors": [
{
"first": "T",
"middle": [],
"last": "Joachims",
"suffix": ""
}
],
"year": 1999,
"venue": "Advances in Kernel Methods -Support Vector Learning",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "T. Joachims. 1999. Making Large-Scale SVM Learn- ing Practical. In B. Sch\u00f6lkopf, C. J.C. Burges, and A. Smola, editors, Advances in Kernel Methods - Support Vector Learning, Cambridge, MA, USA. MIT Press.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Control and initiative in collaborative problem solving dialogues",
"authors": [
{
"first": "P",
"middle": [
"W"
],
"last": "Jordan",
"suffix": ""
},
{
"first": "B",
"middle": [
"Di"
],
"last": "Eugenio",
"suffix": ""
}
],
"year": 1997,
"venue": "Working Notes of the AAAI Spring Symposium on Comp. Models for Mixed Initiative",
"volume": "",
"issue": "",
"pages": "81--84",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "P. W. Jordan and B. Di Eugenio. 1997. Control and initiative in collaborative problem solving dialogues. In Working Notes of the AAAI Spring Symposium on Comp. Models for Mixed Initiative, pages 81-84.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "A Note on Platt's Probabilistic Outputs for Support Vector Machines",
"authors": [
{
"first": "H",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Lin",
"suffix": ""
},
{
"first": "R",
"middle": [
"C"
],
"last": "Weng",
"suffix": ""
}
],
"year": 2007,
"venue": "Mach. Learn",
"volume": "68",
"issue": "",
"pages": "267--276",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "H. Lin, C. Lin, and R. C. Weng. 2007. A Note on Platt's Probabilistic Outputs for Support Vector Ma- chines. Mach. Learn., 68:267-276, October.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Power and politeness in action: disagreements in oral communication. Language, power, and social process",
"authors": [
{
"first": "M",
"middle": [
"A"
],
"last": "Locher",
"suffix": ""
},
{
"first": "; M",
"middle": [],
"last": "De Gruyter",
"suffix": ""
}
],
"year": 2004,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. A. Locher. 2004. Power and politeness in action: disagreements in oral communication. Language, power, and social process. M. de Gruyter.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Gaining turns and achieving high in influence ranking in small conversational groups",
"authors": [
{
"first": "S",
"middle": [
"H"
],
"last": "Ng",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Bell",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Brooke",
"suffix": ""
}
],
"year": 1993,
"venue": "British Journal of Social Psychology",
"volume": "32",
"issue": "",
"pages": "265--275",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "S. H. Ng, D. Bell, and M. Brooke. 1993. Gain- ing turns and achieving high in influence ranking in small conversational groups. British Journal of So- cial Psychology, pages 32, 265-275.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Interruption and in influence in discussion groups",
"authors": [
{
"first": "S",
"middle": [
"H"
],
"last": "Ng",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Brooke",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Dunne",
"suffix": ""
}
],
"year": 1995,
"venue": "Journal of Language and Social Psychology",
"volume": "",
"issue": "4",
"pages": "369--381",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "S. H. Ng, M Brooke, , and M. Dunne. 1995. In- terruption and in influence in discussion groups. Journal of Language and Social Psychology, pages 14(4),369-381.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Linguistic evidence: language, power, and strategy in the courtroom. Studies on law and social control",
"authors": [
{
"first": "W",
"middle": [
"M"
],
"last": "O'barr",
"suffix": ""
}
],
"year": 1982,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "W. M. O'Barr. 1982. Linguistic evidence: language, power, and strategy in the courtroom. Studies on law and social control. Academic Press.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "ClearTK: A UIMA toolkit for statistical natural language processing",
"authors": [
{
"first": "P",
"middle": [
"V"
],
"last": "Ogren",
"suffix": ""
},
{
"first": "P",
"middle": [
"G"
],
"last": "Wetzler",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Bethard",
"suffix": ""
}
],
"year": 2008,
"venue": "Towards Enhanced Interoperability for Large HLT Systems: UIMA for NLP workshop at Language Resources and Evaluation Conference (LREC)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "P. V. Ogren, P. G. Wetzler, and S. Bethard. 2008. ClearTK: A UIMA toolkit for statistical natural lan- guage processing. In Towards Enhanced Interop- erability for Large HLT Systems: UIMA for NLP workshop at Language Resources and Evaluation Conference (LREC).",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Annotations for power relations on email threads",
"authors": [
{
"first": "V",
"middle": [],
"last": "Prabhakaran",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Diab",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the Eighth conference on LREC",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "V. Prabhakaran, O. Rambow, and M. Diab. 2012a. An- notations for power relations on email threads. In Proceedings of the Eighth conference on LREC, Is- tanbul, Turkey, May. ELRA.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Predicting Overt Display of Power in Written Dialogs",
"authors": [
{
"first": "V",
"middle": [],
"last": "Prabhakaran",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Diab",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the HLT-NAACL",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "V. Prabhakaran, O. Rambow, and M. Diab. 2012b. Predicting Overt Display of Power in Written Di- alogs. In Proceedings of the HLT-NAACL, Montreal, Canada, June. ACL.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Who's (Really) the Boss? Perception of Situational Power in Written Interactions",
"authors": [
{
"first": "V",
"middle": [],
"last": "Prabhakaran",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Rambow",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Diab",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 24th International Conference on COLING, Mumbai",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "V. Prabhakaran, O. Rambow, and M. Diab. 2012c. Who's (Really) the Boss? Perception of Situational Power in Written Interactions. In Proceedings of the 24th International Conference on COLING, Mum- bai, India. ACL.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "Conversation as a resource for in influence: evidence for prototypical arguments and social identification processes",
"authors": [
{
"first": "S",
"middle": [
"A"
],
"last": "Reid",
"suffix": ""
},
{
"first": "S",
"middle": [
"H"
],
"last": "Ng",
"suffix": ""
}
],
"year": 2000,
"venue": "European Journal of Social Psych",
"volume": "",
"issue": "",
"pages": "83--100",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "S. A. Reid and S. H. Ng. 2000. Conversation as a resource for in influence: evidence for prototypical arguments and social identification processes. Euro- pean Journal of Social Psych., pages 30, 83-100.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Automated social hierarchy detection through email network analysis",
"authors": [
{
"first": "R",
"middle": [],
"last": "Rowe",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Creamer",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Hershkop",
"suffix": ""
},
{
"first": "S",
"middle": [
"J"
],
"last": "Stolfo",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 9th WebKDD and 1st SNA-KDD 2007 workshop on Web Mining and Social Network Anal",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "R. Rowe, G. Creamer, S. Hershkop, and S.J. Stolfo. 2007. Automated social hierarchy detection through email network analysis. In Proceedings of the 9th WebKDD and 1st SNA-KDD 2007 workshop on Web Mining and Social Network Anal. ACM.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Mixed initiative in dialogue: An investigation into discourse segmentation",
"authors": [
{
"first": "M",
"middle": [],
"last": "Walker",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Whittaker",
"suffix": ""
}
],
"year": 1990,
"venue": "Proceedings of the 28th annual meeting on ACL",
"volume": "",
"issue": "",
"pages": "70--78",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "M. Walker and S. Whittaker. 1990. Mixed initiative in dialogue: An investigation into discourse segmenta- tion. In Proceedings of the 28th annual meeting on ACL, pages 70-78. ACL.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "The forms of power: from domination to transformation",
"authors": [
{
"first": "T",
"middle": [
"E"
],
"last": "Wartenberg",
"suffix": ""
}
],
"year": 1990,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "T. E. Wartenberg. 1990. The forms of power: from domination to transformation. Temple Univ. Press.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "More accurate tests for the statistical significance of result differences",
"authors": [
{
"first": "A",
"middle": [],
"last": "Yeh",
"suffix": ""
}
],
"year": 2000,
"venue": "Proceedings of the 18th conference on COLING",
"volume": "2",
"issue": "",
"pages": "947--953",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A. Yeh. 2000. More accurate tests for the statistical significance of result differences. In Proceedings of the 18th conference on COLING -Volume 2, pages 947-953, Stroudsburg, PA, USA. ACL.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"text": "5: Variations in manifestations of power on feature values: mean(P)| mean(N ) p\u2212value P: people judged to have power; N : people judged not to have power; Values with p \u2264 0.05 are boldfaced Types of power -SP: Situational power, HP: Hierarchical power, PC: Power over Communication, INFL: Influence; Features -DAP: Dialog acts, DLC: Dialog links, PST: Positional, VRB: Verbosity, ODP: Overt display of power P and N .",
"num": null,
"type_str": "figure",
"uris": null
},
"FIGREF1": {
"text": "SP: Situational power, HP: Hierarchical power PC: Power over communication, INFL: Influence VRB: Verbosity, PST: Positional, DAP: Dialog acts, DLC: Dialog links, LEX: Lexical, ODP: Overt display of power lated using Lin et al. (2007)'s algorithm.",
"num": null,
"type_str": "figure",
"uris": null
},
"TABREF1": {
"text": "Power over Communication (PC): A person is said to have power over communication if he actively attempts to achieve the intended goals of the communication. 2 These are people who ask questions, request others to take action, etc., and not From: Kathryn Cordes To: Leslie Hansen, Sara Shackleton, Brent Hendry CC: Mark Greenberg, Erik Eller, Thomas D Gros ----------------M1.1. Leslie Sara, and Brent: [Conventional] M1.2. Could I get an update on were we are with the top 20 customer amendments. [Req-Info]; [Flink1.2] . I have just finished the draft for our internal legal review and sent it around. [Inform]M2.2. There are still a lot of work to be done but I do not know when everyone will have time to look at this considering how much other work there is.[Inform]",
"html": null,
"type_str": "table",
"content": "<table><tr><td colspan=\"2\">M1.3. Last week we got 5 amendments for power</td></tr><tr><td colspan=\"2\">physical but we still haven't received any amendments</td></tr><tr><td>for financial. [Inform]</td><td/></tr><tr><td colspan=\"2\">M1.4. Entergy-Koch is very interested in ConfirmLogic</td></tr><tr><td colspan=\"2\">and have asked for the amendments. [Inform]</td></tr><tr><td colspan=\"2\">M1.5. When can we get the amendments for Entergy-</td></tr><tr><td colspan=\"2\">Koch completed? [Req-Info]; [Flink1.5]</td></tr><tr><td colspan=\"2\">M1.6. Thanks, [Conventional]</td></tr><tr><td>M1.7. KC [Conventional]</td><td/></tr><tr><td>From: Brent Hendry</td><td/></tr><tr><td>To: Mark Taylor</td><td/></tr><tr><td colspan=\"2\">----------------</td></tr><tr><td colspan=\"2\">M2.1M2.3. How should we respond considering she has</td></tr><tr><td colspan=\"2\">copied Tom Gros? [Req-Info]; [Blink1.5]; [Flink2.3]</td></tr><tr><td>Person with SP</td><td>Kathryn Cordes</td></tr><tr><td>Person with PC</td><td>N/A</td></tr><tr><td>Person with INFL</td><td>Mark Taylor</td></tr><tr><td colspan=\"2\">Overt Display of Power M1.2</td></tr></table>",
"num": null
},
"TABREF2": {
"text": "Round 2.",
"html": null,
"type_str": "table",
"content": "<table><tr><td>Type of power</td><td colspan=\"2\">Round 1 Round 2</td></tr><tr><td>Situational Power (SP)</td><td>0.47</td><td>0.47</td></tr><tr><td>Power over Communication (PC)</td><td>0.27</td><td>0.76</td></tr><tr><td>Influence (INFL)</td><td>0.50</td><td>0.79</td></tr></table>",
"num": null
},
"TABREF3": {
"text": "",
"html": null,
"type_str": "table",
"content": "<table/>",
"num": null
},
"TABREF5": {
"text": "",
"html": null,
"type_str": "table",
"content": "<table/>",
"num": null
},
"TABREF7": {
"text": "Cross validation results",
"html": null,
"type_str": "table",
"content": "<table/>",
"num": null
}
}
}
} |