基於常識知識的移情對話回覆生成

Tzu Hsien Huang; Chia Hui Chang

基於常識知識的移情對話回覆生成

Translated title of the contribution: Improving Response Diversity through Commonsense-Aware Empathetic Response Generation

Tzu Hsien Huang, Chia Hui Chang

Department of Computer Science and Information Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Due to the lack of conversation practice, the main challenge for the second-language learners is speaking. Our goal is to develop a chatbot to encourage individuals to reflect, describe, analyse and communicate what they read as well as improve students' English expression skills. In this paper, we exploit COMMET, an inferential commonsense knowledge generator, as the background knowledge to improve the generation diversity. We consider two approaches to increase the diversity of empathetic response generation. For non-pretrained models, We apply AdaLabel (Wang et al., 2021) to Commonsense-aware Empathetic model (Sabour et al., 2022) and improve Distinct-2 score from 2.99 to 4.08 on EMPATHETIC DIALOGUES (ED). Furthermore, we augment the pretrained BART model with various commonsense knowledge to generate more informative empathetic responses. Not only has the automatic evaluation of distinct-2 scores improved from 9.11 to 11.21, but the manual case study also shows that CE-BART significantly outperform CEM-AdaLabel.

Translated title of the contribution	Improving Response Diversity through Commonsense-Aware Empathetic Response Generation
Original language	Chinese (Traditional)
Title of host publication	ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing
Editors	Yung-Chun Chang, Yi-Chin Huang, Jheng-Long Wu, Ming-Hsiang Su, Hen-Hsen Huang, Yi-Fen Liu, Lung-Hao Lee, Chin-Hung Chou, Yuan-Fu Liao
Publisher	The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)
Pages	299-306
Number of pages	8
ISBN (Electronic)	9789869576956
State	Published - 2022
Event	34th Conference on Computational Linguistics and Speech Processing, ROCLING 2022 - Taipei, Taiwan Duration: 21 Nov 2022 → 22 Nov 2022

Publication series

Name	ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing

Conference

Conference	34th Conference on Computational Linguistics and Speech Processing, ROCLING 2022
Country/Territory	Taiwan
City	Taipei
Period	21/11/22 → 22/11/22

Cite this

Huang, T. H., & Chang, C. H. (2022). 基於常識知識的移情對話回覆生成. In Y.-C. Chang, Y.-C. Huang, J.-L. Wu, M.-H. Su, H.-H. Huang, Y.-F. Liu, L.-H. Lee, C.-H. Chou, & Y.-F. Liao (Eds.), ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing (pp. 299-306). (ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing). The Association for Computational Linguistics and Chinese Language Processing (ACLCLP).

Huang, Tzu Hsien ; Chang, Chia Hui. / 基於常識知識的移情對話回覆生成. ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing. editor / Yung-Chun Chang ; Yi-Chin Huang ; Jheng-Long Wu ; Ming-Hsiang Su ; Hen-Hsen Huang ; Yi-Fen Liu ; Lung-Hao Lee ; Chin-Hung Chou ; Yuan-Fu Liao. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), 2022. pp. 299-306 (ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing).

@inproceedings{a68c322621e641cc8c65835117f3862d,

title = "基於常識知識的移情對話回覆生成",

abstract = "Due to the lack of conversation practice, the main challenge for the second-language learners is speaking. Our goal is to develop a chatbot to encourage individuals to reflect, describe, analyse and communicate what they read as well as improve students' English expression skills. In this paper, we exploit COMMET, an inferential commonsense knowledge generator, as the background knowledge to improve the generation diversity. We consider two approaches to increase the diversity of empathetic response generation. For non-pretrained models, We apply AdaLabel (Wang et al., 2021) to Commonsense-aware Empathetic model (Sabour et al., 2022) and improve Distinct-2 score from 2.99 to 4.08 on EMPATHETIC DIALOGUES (ED). Furthermore, we augment the pretrained BART model with various commonsense knowledge to generate more informative empathetic responses. Not only has the automatic evaluation of distinct-2 scores improved from 9.11 to 11.21, but the manual case study also shows that CE-BART significantly outperform CEM-AdaLabel.",

keywords = "Commonsense aware response generation, Empathetic response generation, response diversity",

author = "Huang, {Tzu Hsien} and Chang, {Chia Hui}",

note = "Publisher Copyright: {\textcopyright} 2022 the Association for Computational Linguistics and Chinese Language Processing (ACLCLP).; 34th Conference on Computational Linguistics and Speech Processing, ROCLING 2022 ; Conference date: 21-11-2022 Through 22-11-2022",

year = "2022",

language = "繁體中文",

series = "ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing",

publisher = "The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)",

pages = "299--306",

editor = "Yung-Chun Chang and Yi-Chin Huang and Jheng-Long Wu and Ming-Hsiang Su and Hen-Hsen Huang and Yi-Fen Liu and Lung-Hao Lee and Chin-Hung Chou and Yuan-Fu Liao",

booktitle = "ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing",

}

Huang, TH & Chang, CH 2022, 基於常識知識的移情對話回覆生成. in Y-C Chang, Y-C Huang, J-L Wu, M-H Su, H-H Huang, Y-F Liu, L-H Lee, C-H Chou & Y-F Liao (eds), ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing. ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing, The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), pp. 299-306, 34th Conference on Computational Linguistics and Speech Processing, ROCLING 2022, Taipei, Taiwan, 21/11/22.

基於常識知識的移情對話回覆生成. / Huang, Tzu Hsien; Chang, Chia Hui.
ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing. ed. / Yung-Chun Chang; Yi-Chin Huang; Jheng-Long Wu; Ming-Hsiang Su; Hen-Hsen Huang; Yi-Fen Liu; Lung-Hao Lee; Chin-Hung Chou; Yuan-Fu Liao. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), 2022. p. 299-306 (ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - 基於常識知識的移情對話回覆生成

AU - Huang, Tzu Hsien

AU - Chang, Chia Hui

PY - 2022

Y1 - 2022

N2 - Due to the lack of conversation practice, the main challenge for the second-language learners is speaking. Our goal is to develop a chatbot to encourage individuals to reflect, describe, analyse and communicate what they read as well as improve students' English expression skills. In this paper, we exploit COMMET, an inferential commonsense knowledge generator, as the background knowledge to improve the generation diversity. We consider two approaches to increase the diversity of empathetic response generation. For non-pretrained models, We apply AdaLabel (Wang et al., 2021) to Commonsense-aware Empathetic model (Sabour et al., 2022) and improve Distinct-2 score from 2.99 to 4.08 on EMPATHETIC DIALOGUES (ED). Furthermore, we augment the pretrained BART model with various commonsense knowledge to generate more informative empathetic responses. Not only has the automatic evaluation of distinct-2 scores improved from 9.11 to 11.21, but the manual case study also shows that CE-BART significantly outperform CEM-AdaLabel.

AB - Due to the lack of conversation practice, the main challenge for the second-language learners is speaking. Our goal is to develop a chatbot to encourage individuals to reflect, describe, analyse and communicate what they read as well as improve students' English expression skills. In this paper, we exploit COMMET, an inferential commonsense knowledge generator, as the background knowledge to improve the generation diversity. We consider two approaches to increase the diversity of empathetic response generation. For non-pretrained models, We apply AdaLabel (Wang et al., 2021) to Commonsense-aware Empathetic model (Sabour et al., 2022) and improve Distinct-2 score from 2.99 to 4.08 on EMPATHETIC DIALOGUES (ED). Furthermore, we augment the pretrained BART model with various commonsense knowledge to generate more informative empathetic responses. Not only has the automatic evaluation of distinct-2 scores improved from 9.11 to 11.21, but the manual case study also shows that CE-BART significantly outperform CEM-AdaLabel.

KW - Commonsense aware response generation

KW - Empathetic response generation

KW - response diversity

UR - http://www.scopus.com/inward/record.url?scp=85154584258&partnerID=8YFLogxK

M3 - 會議論文篇章

AN - SCOPUS:85154584258

T3 - ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing

SP - 299

EP - 306

BT - ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing

A2 - Chang, Yung-Chun

A2 - Huang, Yi-Chin

A2 - Wu, Jheng-Long

A2 - Su, Ming-Hsiang

A2 - Huang, Hen-Hsen

A2 - Liu, Yi-Fen

A2 - Lee, Lung-Hao

A2 - Chou, Chin-Hung

A2 - Liao, Yuan-Fu

PB - The Association for Computational Linguistics and Chinese Language Processing (ACLCLP)

T2 - 34th Conference on Computational Linguistics and Speech Processing, ROCLING 2022

Y2 - 21 November 2022 through 22 November 2022

ER -

Huang TH, Chang CH. 基於常識知識的移情對話回覆生成. In Chang YC, Huang YC, Wu JL, Su MH, Huang HH, Liu YF, Lee LH, Chou CH, Liao YF, editors, ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing. The Association for Computational Linguistics and Chinese Language Processing (ACLCLP). 2022. p. 299-306. (ROCLING 2022 - Proceedings of the 34th Conference on Computational Linguistics and Speech Processing).

基於常識知識的移情對話回覆生成

Abstract

Publication series

Conference

Fingerprint

Cite this