Beyond Candidates: Adaptive Dialogue Agent Utilizing Persona and Knowledge

Jungwoo Lim, Myunghoon Kang, Jinsung Kim, Jeongwook Kim, Yuna Hur, Heuiseok Lim

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

To build ultimate dialogue agents, previous studies suggest models that ground both persona and knowledge. However, applying the dialogue system directly to the usual conversation is still limited because the system requires a complete sentence-formed persona and knowledge candidate sets from the given dataset. In contrast to the dialogue setting in the dataset, humans utilize semantic concepts in their minds rather than a set of predefined candidate sentences. Following this manner of human dialogue, we suggest an adaptive dialogue system that is applicable to situations where complete sentence-formed candidates are not given. Our model generates consistent and relevant persona descriptions and identifies relevant knowledge for engaging and knowledgeable responses, even with fragmentary information. We show that our model outperforms previous baselines that utilize persona and knowledge candidate sentences and conduct the human evaluation on the machine-generated responses. In addition, we conduct ablation studies to demonstrate the effectiveness of each component of our model. Furthermore, we apply our model to other dialogue datasets that only ground knowledge or persona to showcase its adaptability. Our code is available at https://github.com/dlawjddn803/BeCand.

Original languageEnglish
Title of host publicationFindings of the Association for Computational Linguistics
Subtitle of host publicationEMNLP 2023
PublisherAssociation for Computational Linguistics (ACL)
Pages7950-7963
Number of pages14
ISBN (Electronic)9798891760615
Publication statusPublished - 2023
Event2023 Findings of the Association for Computational Linguistics: EMNLP 2023 - Singapore, Singapore
Duration: 2023 Dec 62023 Dec 10

Publication series

NameFindings of the Association for Computational Linguistics: EMNLP 2023

Conference

Conference2023 Findings of the Association for Computational Linguistics: EMNLP 2023
Country/TerritorySingapore
CitySingapore
Period23/12/623/12/10

Bibliographical note

Publisher Copyright:
© 2023 Association for Computational Linguistics.

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Science Applications
  • Information Systems
  • Language and Linguistics
  • Linguistics and Language

Fingerprint

Dive into the research topics of 'Beyond Candidates: Adaptive Dialogue Agent Utilizing Persona and Knowledge'. Together they form a unique fingerprint.

Cite this