English

Dual Task Framework for Improving Persona-grounded Dialogue Dataset

Computation and Language 2022-02-17 v2 Artificial Intelligence Machine Learning

Abstract

This paper introduces a simple yet effective data-centric approach for the task of improving persona-conditioned dialogue agents. Prior model-centric approaches unquestioningly depend on the raw crowdsourced benchmark datasets such as Persona-Chat. In contrast, we aim to fix annotation artifacts in benchmarking, which is orthogonally applicable to any dialogue model. Specifically, we augment relevant personas to improve dialogue dataset/agent, by leveraging the primal-dual structure of the two tasks, predicting dialogue responses and personas based on each other. Experiments on Persona-Chat show that our approach outperforms pre-trained LMs by an 11.7 point gain in terms of accuracy.

Keywords

Cite

@article{arxiv.2202.05435,
  title  = {Dual Task Framework for Improving Persona-grounded Dialogue Dataset},
  author = {Minju Kim and Beong-woo Kwak and Youngwook Kim and Hong-in Lee and Seung-won Hwang and Jinyoung Yeo},
  journal= {arXiv preprint arXiv:2202.05435},
  year   = {2022}
}

Comments

Accepted to AAAI2022

R2 v1 2026-06-24T09:31:26.363Z