TY - GEN
T1 - Polite Task-oriented Dialog Agents
T2 - 12th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, WASSA 2022
AU - Silva, Diogo
AU - Semedo, David
AU - Magalhães, João
N1 - info:eu-repo/grantAgreement/FCT/OE/PRT%2FBD%2F152810%2F2021/PT#
info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F04516%2F2020/PT#
Funding Information:
This work has been partially funded by the iFetch project, Ref. 45920, co-financed by ERDF, COMPETE 2020, NORTE 2020 and FCT under CMU Portugal.
Publisher Copyright:
© 2022 Association for Computational Linguistics.
PY - 2022
Y1 - 2022
N2 - For task-oriented dialog agents, the tone of voice mediates user-agent interactions, playing a central role in the flow of a conversation. Distinct from domain-agnostic politeness constructs, in specific domains such as online stores, booking platforms, and others, agents need to be capable of adopting highly specific vocabulary, with significant impact on lexical and grammatical aspects of utterances. Then, the challenge is on improving utterances’ politeness while preserving the actual content, an utterly central requirement to achieve the task goal. In this paper, we conduct a novel assessment of politeness strategies for task-oriented dialog agents under a transfer learning scenario. We extend existing generative and rewriting politeness approaches, towards overcoming domain-shifting issues, and enabling the transfer of politeness patterns to a novel domain. Both automatic and human evaluation is conducted on customer-store interactions, over the fashion domain, from which contribute with insightful and experimentally supported lessons regarding the improvement of politeness in task-specific dialog agents.
AB - For task-oriented dialog agents, the tone of voice mediates user-agent interactions, playing a central role in the flow of a conversation. Distinct from domain-agnostic politeness constructs, in specific domains such as online stores, booking platforms, and others, agents need to be capable of adopting highly specific vocabulary, with significant impact on lexical and grammatical aspects of utterances. Then, the challenge is on improving utterances’ politeness while preserving the actual content, an utterly central requirement to achieve the task goal. In this paper, we conduct a novel assessment of politeness strategies for task-oriented dialog agents under a transfer learning scenario. We extend existing generative and rewriting politeness approaches, towards overcoming domain-shifting issues, and enabling the transfer of politeness patterns to a novel domain. Both automatic and human evaluation is conducted on customer-store interactions, over the fashion domain, from which contribute with insightful and experimentally supported lessons regarding the improvement of politeness in task-specific dialog agents.
UR - http://www.scopus.com/inward/record.url?scp=85137711002&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85137711002
T3 - WASSA 2022 - 12th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, Proceedings of the Workshop
SP - 304
EP - 314
BT - WASSA 2022
A2 - Barnes, Jeremy
A2 - De Clercq, Orphee
A2 - Barriere, Valentin
A2 - Tafreshi, Shabnam
A2 - Alqahtani, Sawsan
A2 - Sedoc, Joao
A2 - Klinger, Roman
A2 - Balahur, Alexandra
PB - Association for Computational Linguistics (ACL)
Y2 - 26 May 2022
ER -