File size: 37,898 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 |
{
"paper_id": "W07-0209",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T04:40:10.924097Z"
},
"title": "Semi-supervised Algorithm for Human-Computer Dialogue Mining",
"authors": [
{
"first": "Calkin",
"middle": [
"S"
],
"last": "Montero",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Hokkaido University",
"location": {
"addrLine": "Kita 14-jo Nishi 9-chome Kita-ku",
"postCode": "060-0814",
"settlement": "Sapporo",
"country": "Japan"
}
},
"email": "[email protected]"
},
{
"first": "Kenji",
"middle": [],
"last": "Araki",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Hokkaido University",
"location": {
"addrLine": "Kita 14-jo Nishi 9-chome Kita-ku",
"postCode": "060-0814",
"settlement": "Sapporo",
"country": "Japan"
}
},
"email": "[email protected]"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper describes the analysis of weak local coherence utterances during humancomputer conversation through the application of an emergent data mining technique, data crystallization. Results reveal that by adding utterances with weak local relevance the performance of a baseline conversational partner, in terms of user satisfaction, showed betterment.",
"pdf_parse": {
"paper_id": "W07-0209",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper describes the analysis of weak local coherence utterances during humancomputer conversation through the application of an emergent data mining technique, data crystallization. Results reveal that by adding utterances with weak local relevance the performance of a baseline conversational partner, in terms of user satisfaction, showed betterment.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "Data mining can be defined as the process of finding new and potentially useful knowledge from data. An enhanced trend of data mining is chance discovery, which in spite of being an emergent field of research has been applied to different branches of science. Recently, data crystallization (Ohsawa, 2005) has been proposed as a chance discovery extension devoted to find desired unobservable events within a given data. This method of data mining has been intended for revealing events that are significant but are not included in the analyzed data (sparseness).",
"cite_spans": [
{
"start": 291,
"end": 305,
"text": "(Ohsawa, 2005)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The knowledge database of a conversational partner computer program is a good example of data sparseness, making very difficult to realize a humanlike conversation between a user and a computer, since the database is missing relevant or coherent answers for a given user input, e.g., ELIZA system (Weizenbaum, 1966) , corpus-based dialogue systems (Inui et al., 2003) , and so forth. Herring (1999) noticed the weak sequential coherence during human-computer interaction that is caused by weak local relevance, pointing out the violation of the Gricean maxim of local relevance during the dialogue. However, we argue that conversational responses -often of very weak local relevance -generated by a conversational partner computer program (or \"chatbot\") could successfully simulate humanlike coherent dialogue (see example given by (Herring, 1999) , where topical coherence is stated in spite of violation of local relevance Gricean maxim). Moreover, we argue that weak local relevance utterances could be regarded as topic-shifting utterances. These utterances are vital in order to keep the flow of the dialogue and its overall coherence. Determining the moment when the topic-shifting utterances are needed is hence essential to the success of a chatbot.",
"cite_spans": [
{
"start": 297,
"end": 315,
"text": "(Weizenbaum, 1966)",
"ref_id": "BIBREF15"
},
{
"start": 348,
"end": 367,
"text": "(Inui et al., 2003)",
"ref_id": "BIBREF11"
},
{
"start": 384,
"end": 398,
"text": "Herring (1999)",
"ref_id": "BIBREF10"
},
{
"start": 832,
"end": 847,
"text": "(Herring, 1999)",
"ref_id": "BIBREF10"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "In this paper KeyGraph and data crystallization techniques are applied to analyze human-computer trivial dialogue, chat, considering as chances to discover the missing topic-shifting utterances, used to improve the overall sequential coherence and the flow of the computer dialogue.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "During the process of dialogue mining, in order to analyze the behavior of human-computer conversational interaction, the interrelationship of utterances of a dialogue section was obtained. The following is a fragment of an analyzed dialogue section between a user and ALICE chatbot (Wallace, 2005 For the analysis, the application of data crystallization is introduced in order to identify the missing topic-shifting utterances. The interrelationship of utterances (user-chatbot) is represented in a cooccurrence document created as described below.",
"cite_spans": [
{
"start": 283,
"end": 297,
"text": "(Wallace, 2005",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Human-Computer Conversation Analysis Approach",
"sec_num": "2"
},
{
"text": "The KeyGraph has been used as a data-mining tool for extracting patterns of the appearance of chance events (Ohsawa et al.(2003) ). The KeyGraph identifies relationships between terms in a document particularly focusing on co-occurrence relationships of both high-probability and low-probability events. Montero et al. (2005) have applied this tool for analyzing the dynamic behavior of human-human chat, identifying criticality.",
"cite_spans": [
{
"start": 108,
"end": 128,
"text": "(Ohsawa et al.(2003)",
"ref_id": null
},
{
"start": 319,
"end": 325,
"text": "(2005)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "The KeyGraph",
"sec_num": "2.1"
},
{
"text": "In this paper the KeyGraph is applied in combination with data crystallization in order to visualized utterances that do not appear during humancomputer chat. The interrelationship of utterances (user-chatbot) is represented in a co-occurrence document created by the following algorithm: a) Each utterance (from both, the user and the chatbot) was considered as Then it could be said that the obtained D document contains the co-occurrence relationship of the utterances during the analyzed dialogue section. In the graph, the most frequent items in D are shown as black nodes and the most strongly co-occurring item-pairs are linked by black lines according to the Jaccard coefficient:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "The KeyGraph",
"sec_num": "2.1"
},
{
"text": "J(S x , S y ) = p(S x \u2229 S y )/p(S x \u222a S y )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "The KeyGraph",
"sec_num": "2.1"
},
{
"text": "where p(S x \u2229 S y ) is the probability that both elements S x and S y co-occur in a line in D, and p(S x \u222a S y ) is the probability that either S x or S y appears in a line. In the graph, nodes are interpreted as sentences (from D) and clusters of nodes as particular topics (Figure 1 ).",
"cite_spans": [],
"ref_spans": [
{
"start": 275,
"end": 284,
"text": "(Figure 1",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "The KeyGraph",
"sec_num": "2.1"
},
{
"text": "Data crystallization (Ohsawa, 2005) , is dedicated to experts working in real domains where discoveries of events that are important but are not included in the analyzed data are desired. The process of data crystallization involves to insert dummy items in the given data in order to represent unobservable events. In this paper, each dummy item inserted in the D document (one in each vector of D) is named X Y , where X represents the level of the insertion and Y represents the line where the dummy item was inserted. The KeyGraph is applied to the new D document and all of the dummy nodes that did not appear linking clusters in the graph are eliminated from the data, and then the cycle is iterated to higher levels. In the case of the D document of Sec.2.1, after the first level of insertion it becomes: After feeding the KeyGraph with D \u2032 , all the dummy items that did not appear linking clusters as bridges in the outputted graph are deleted. At this point new dummy items with higher hierarchy (2 x ) are inserted in D \u2032 , and the cycle iterates. Unobservable events and their relations with other events are to be visualized by the application of KeyGraph iteratively to the data that is been crystallized (Figure 2 ). ",
"cite_spans": [
{
"start": 21,
"end": 35,
"text": "(Ohsawa, 2005)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [
{
"start": 1220,
"end": 1229,
"text": "(Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Data Crystallization",
"sec_num": "2.2"
},
{
"text": "D \u2032 =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Crystallization",
"sec_num": "2.2"
},
{
"text": "The performed experiment was carried out in three stages. In the first stage of the experiment, three different dialogue sections (including the one shown in Sec.2) between three native English speakers and a chatbot (Wallace, 2005) were analyzed in order to find co-occurrence between the users' utterance and the chatbot replies, i.e., D document. This D document was then examined by the KeyGraph (unsupervised process). Figure 1 shows the graphical view of the dialogue in Sec.2 (48 turns, user -chatbot, in total). A characteristic of the KeyGraph is the visualization of co-occurring events by means of clusters. In Figure 1 , the nodes represent sentences from the D document, the clusters represent the relationship among those sentences, i.e., a specific topic, and the nodes that link the clusters represent the transition from one topic to the next. It can be observed that the main clusters are not interconnected, leading to the conclusion that the chatbot in many cases could not keep a smooth and natural flow of the dialogue.",
"cite_spans": [
{
"start": 217,
"end": 232,
"text": "(Wallace, 2005)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [
{
"start": 424,
"end": 432,
"text": "Figure 1",
"ref_id": "FIGREF2"
},
{
"start": 622,
"end": 630,
"text": "Figure 1",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "In the second stage of the experiment, a crystallized document of utterance co-occurrence, i.e., D \u2032 document, was obtained for the same dialogue sections, following the process described in Sec.2.2. The graphical output of the dialogue in Sec.2, after crystallization, can be observed in Figure 2 . It can be seen in this figure how the two main clusters appear to be interconnected by the dummy item 1 3.",
"cite_spans": [],
"ref_spans": [
{
"start": 289,
"end": 297,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "Although this dummy item was inserted in the third line of the D document, it appears in the graph connecting the two main clusters. The dummy item 1 3 branches from utterance [24] . This interconnecting point can be regarded as the system considering it appropriate to insert a topic-shifting utterance at this point of the conversation. In doing so, a well interconnected graph is obtained (Figure 2 ). This information is valuable for making the chatbot to ask \"intelligent questions\" as a mean of conversational responses to keep the interest from the user.",
"cite_spans": [
{
"start": 176,
"end": 180,
"text": "[24]",
"ref_id": "BIBREF9"
}
],
"ref_spans": [
{
"start": 392,
"end": 401,
"text": "(Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "In the third stage of the experiment, the information yielded by the previous analysis, i.e., regarding the timing where a topic-shifting utterance might be needed, was used to feed the chatbot database. Topic-shifting responses were inserted by hand (supervised process) as general patterns (around one hundred patterns) for smoothly change the topic when there is not a pattern that matches a given utterance. In this way a bridge, represented in Figure 2 by the dummy item, is created giving to the dialogue the desired smoothness. Seven users (four native English speakers, three non native speakers) were requested to perform a chat with the plain chatbot and with the enhanced chatbot (the users did not know which chatbot was plain or which was enhanced). The time set up was maximum 30 minuteschat with each program, the user was free to stop at any time before the time limit. The evaluation of the chatbots performances was made through a ques- The four native English speaker users globally agreed ranking enhanced chatbot as having \"good\" accurate/relevant responses during the overall dialogue, giving remarks like \"interesting to talk with\", \"better at keeping the conversation going\", \"easier to talk to\", and \"more impression of consciousness\". In the case of the plain chatbot, the native speakers showed \"dissatisfied\" at its performance, giving remarks like \"loop responses\", \"slow in keeping the conversation going\", \"no so nice responses\", \"no constant memory\". Table 1 shows a resume of the average performance of the chatbot for all of the users. An example of a vague reply is given in the following fragment: Two non native English speaker users ranked the enhanced chatbot as having \"fair\" and \"average\" accurate/relevant responses while the plain chatbot was ranked as having \"poor\" and \"fair\" accurate/relevant responses. The third non native English speaker user ranked both chatbots as \"poor\" due to \"the chatbots lack of understanding deixis, and anaphor\".",
"cite_spans": [],
"ref_spans": [
{
"start": 449,
"end": 458,
"text": "Figure 2",
"ref_id": null
},
{
"start": 1484,
"end": 1491,
"text": "Table 1",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "As a mean of discussion, in Figure 2 it could be expected that the dummy item 1 3 would branch from utterance [25] {User: no, you asked me who is the best robot}, which is in the same cluster with utterance [24] . However, under closer examination it becomes clear that utterance [24] has stronger co-occurrence with utterance [38] {Chatbot: I know you are but what am I} than utterance [25] . Hence, the algorithm suggests to link the clusters via utterance [24] .",
"cite_spans": [
{
"start": 207,
"end": 211,
"text": "[24]",
"ref_id": "BIBREF9"
},
{
"start": 280,
"end": 284,
"text": "[24]",
"ref_id": "BIBREF9"
},
{
"start": 387,
"end": 391,
"text": "[25]",
"ref_id": null
},
{
"start": 459,
"end": 463,
"text": "[24]",
"ref_id": "BIBREF9"
}
],
"ref_spans": [
{
"start": 28,
"end": 36,
"text": "Figure 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "In other aspect, based on the feedback given by the seven users of the experiment, the overall performance of the enhanced chatbot can be considered better than the plain chatbot. It is worth noticing that the evaluation of the non native English speaker users tended to emphasize the grammatical aspect of the chatbots responses. On the other hand, the evaluation of the native English speaker users tended to emphasize the smoothness of the dialogue. Although there is still plenty of room for improvement and research a betterment in the chatbot performance could be seen through this approach.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiment and Visual Results",
"sec_num": "3"
},
{
"text": "In this paper the application of a novel data mining method, data crystallization, for visualizing missing topic-shifting utterances during human-computer chat has been described. Based on this information, during the experiment, the use of weak local relevance utterances, i.e., topic-shifting responses, despite of violation of Grecian maxim of local relevance, showed to meliorate the overall dialogue flow. Future research will be oriented to the extended implementation of the obtained results for enhancing the chat flow modeling of a conversational partner program.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "4"
},
{
"text": "Since follow-ups were eliminated, the number of sentences in D might be smaller than the actual number of sentences in the dialogue.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "User: what should I call you?",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "User: what should I call you?",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Chatbot: my name is Alice",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chatbot: my name is Alice",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "User: what do you mean by robot? there are lots of different robots",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "User: what do you mean by robot? there are lots of differ- ent robots",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Have you tried a web search? Would you like to know more?",
"authors": [
{
"first": "",
"middle": [],
"last": "Chatbot",
"suffix": ""
}
],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chatbot: It is a good question. Have you tried a web search? Would you like to know more?",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "User: sure",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "User: sure",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Chatbot: Ok let's do it",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chatbot: Ok let's do it. What would you like to know?",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "User: what Alice thinks a robot means",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "User: what Alice thinks a robot means",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Chatbot: is that a rhetorical question? I would search for it",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chatbot: is that a rhetorical question? I would search for it",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "User: when I hear robot, I think of something that moves",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "User: when I hear robot, I think of something that moves",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Chatbot: you never know what to expect from life. Why? oh yeah, we are talking about something that moves",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chatbot: you never know what to expect from life. Why? oh yeah, we are talking about something that moves [....] References",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Interactional coherence in cmc",
"authors": [
{
"first": "Susan",
"middle": [],
"last": "Herring",
"suffix": ""
}
],
"year": 1999,
"venue": "Journal of Computer Mediated Communication",
"volume": "4",
"issue": "1",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Susan Herring. 1999. Interactional coherence in cmc. Journal of Computer Mediated Communication, 4(1).",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Fully corpus-based natural language dialogue system",
"authors": [
{
"first": "Nobuo",
"middle": [],
"last": "Inui",
"suffix": ""
},
{
"first": "Takuya",
"middle": [],
"last": "Koiso",
"suffix": ""
},
{
"first": "Junpei",
"middle": [],
"last": "Nakamura",
"suffix": ""
},
{
"first": "Yoshiyuki",
"middle": [],
"last": "Kotani",
"suffix": ""
}
],
"year": 2003,
"venue": "Natural Language Generation in Spoken and Written Dialogue, AAAI Spring Symposium",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nobuo Inui, Takuya Koiso, Junpei Nakamura, and Yoshiyuki Kotani. 2003. Fully corpus-based natural language dia- logue system. In Natural Language Generation in Spoken and Written Dialogue, AAAI Spring Symposium.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Data crystallization: Chance discovery extended for dealing with unobservable events",
"authors": [
{
"first": "Yukio",
"middle": [],
"last": "Ohsawa",
"suffix": ""
}
],
"year": 2005,
"venue": "New Mathematics and Natural Computation",
"volume": "1",
"issue": "3",
"pages": "373--392",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yukio Ohsawa. 2005. Data crystallization: Chance discovery extended for dealing with unobservable events. New Mathe- matics and Natural Computation, 1(3):373-392.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Human chat and selforganized criticality: A chance discovery application",
"authors": [
{
"first": "S",
"middle": [],
"last": "Calkin",
"suffix": ""
},
{
"first": "Kenji",
"middle": [],
"last": "Montero",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Araki",
"suffix": ""
}
],
"year": 2005,
"venue": "New Mathematics and Natural Computation",
"volume": "1",
"issue": "3",
"pages": "407--420",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Calkin S Montero and Kenji Araki. 2005. Human chat and self- organized criticality: A chance discovery application. New Mathematics and Natural Computation, 1(3):407-420.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "A.l.i.c.e. artificial intelligence foundation",
"authors": [
{
"first": "Richard",
"middle": [],
"last": "Wallace",
"suffix": ""
}
],
"year": 2005,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Richard Wallace. 2005. A.l.i.c.e. artificial intelligence founda- tion. http://www.alicebot.org.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Eliza a computer program for the study of natural language communication between man and machine",
"authors": [
{
"first": "Joseph",
"middle": [],
"last": "Weizenbaum",
"suffix": ""
}
],
"year": 1966,
"venue": "Commun. ACM",
"volume": "9",
"issue": "1",
"pages": "36--45",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Joseph Weizenbaum. 1966. Eliza a computer program for the study of natural language communication between man and machine. Commun. ACM, 9(1):36-45.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"num": null,
"uris": null,
"text": "one sentence. b) Each sentence was segmented into words. c) High frequency words were eliminated, i.e., I, you, is, followups and the like, as to avoid false co-occurrence. d) A vectorial representation of each sentence (at word level) was obtained and sentences co-occurrence relationship was determined as 1 : D= w1:: S1, S2, S4 .../ w2:: S9, S25 .../ w3:: S1, S3, S10 .../ ... / wn:: S24, S25, ... Smwhere: w k (k = 1, 2, 3, ..., n), represents a word in a sentence. S l (l = 1, 2, 3, ..., m), represents a sentence."
},
"FIGREF1": {
"type_str": "figure",
"num": null,
"uris": null,
"text": "w1:: S1, S2, S4 ...11 / w2:: S9, S25 ...12 / w3:: S1, S3, S10 ...13 / ... / wn:: S24, S25, ... Sm 1n where 1 o (o = 1, 2, 3, ..., n), represents each dummy item inserted in each vector of D."
},
"FIGREF2": {
"type_str": "figure",
"num": null,
"uris": null,
"text": "User-Computer Chat Graph Figure 2: Crystallized Data Graph"
},
"FIGREF3": {
"type_str": "figure",
"num": null,
"uris": null,
"text": "What is like living in Sapporo? User: sweet as Chatbot: In the context of living in Sapporo I don't understand \"what is like living in Sapporo?\" [vague reply]"
},
"TABREF1": {
"type_str": "table",
"html": null,
"num": null,
"text": ".R.* = Vague Reply (This table contains the average of Turns and VR)",
"content": "<table><tr><td/><td colspan=\"2\">Turns % V.R.*</td><td>Accuracy/relevance</td></tr><tr><td>Baseline</td><td>74</td><td>21.11%</td><td>fair</td></tr><tr><td>Enhanced</td><td>128</td><td>7.37%</td><td>good</td></tr></table>"
},
"TABREF2": {
"type_str": "table",
"html": null,
"num": null,
"text": "Chatbots Overall Performance tionnaire given to the users. In the questionnaire the degree of satisfaction of the user was surveyed based on the accuracy/relevance of the chatbot responses and its ability to keep the dialogue flowing.",
"content": "<table><tr><td>The ac-</td></tr></table>"
}
}
}
} |