Controlling Dialogue Generation with Semantic Exemplars

Prakhar Gupta, Jeffrey Bigham, Yulia Tsvetkov, Amy Pavel


Abstract
Dialogue systems pretrained with large language models generate locally coherent responses, but lack fine-grained control over responses necessary to achieve specific goals. A promising method to control response generation is exemplar-based generation, in which models edit exemplar responses that are retrieved from training data, or hand-written to strategically address discourse-level goals, to fit new dialogue contexts. We present an Exemplar-based Dialogue Generation model, EDGE, that uses the semantic frames present in exemplar responses to guide response generation. We show that controlling dialogue generation based on the semantic frames of exemplars improves the coherence of generated responses, while preserving semantic meaning and conversation goals present in exemplar responses.
Anthology ID:
2021.naacl-main.240
Volume:
Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
Month:
June
Year:
2021
Address:
Online
Editors:
Kristina Toutanova, Anna Rumshisky, Luke Zettlemoyer, Dilek Hakkani-Tur, Iz Beltagy, Steven Bethard, Ryan Cotterell, Tanmoy Chakraborty, Yichao Zhou
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3018–3029
Language:
URL:
https://aclanthology.org/2021.naacl-main.240
DOI:
10.18653/v1/2021.naacl-main.240
Bibkey:
Cite (ACL):
Prakhar Gupta, Jeffrey Bigham, Yulia Tsvetkov, and Amy Pavel. 2021. Controlling Dialogue Generation with Semantic Exemplars. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3018–3029, Online. Association for Computational Linguistics.
Cite (Informal):
Controlling Dialogue Generation with Semantic Exemplars (Gupta et al., NAACL 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.naacl-main.240.pdf
Video:
 https://aclanthology.org/2021.naacl-main.240.mp4
Code
 prakharguptaz/EDGE-exemplars
Data
DailyDialogFrameNet