Dual Task Framework for Improving Persona-Grounded Dialogue Dataset
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.
Cite
Text
Kim et al. "Dual Task Framework for Improving Persona-Grounded Dialogue Dataset." AAAI Conference on Artificial Intelligence, 2022. doi:10.1609/AAAI.V36I10.21338Markdown
[Kim et al. "Dual Task Framework for Improving Persona-Grounded Dialogue Dataset." AAAI Conference on Artificial Intelligence, 2022.](https://mlanthology.org/aaai/2022/kim2022aaai-dual/) doi:10.1609/AAAI.V36I10.21338BibTeX
@inproceedings{kim2022aaai-dual,
title = {{Dual Task Framework for Improving Persona-Grounded Dialogue Dataset}},
author = {Kim, Minju and Kwak, Beong-woo and Kim, Youngwook and Lee, Hong-in and Hwang, Seung-won and Yeo, Jinyoung},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2022},
pages = {10912-10920},
doi = {10.1609/AAAI.V36I10.21338},
url = {https://mlanthology.org/aaai/2022/kim2022aaai-dual/}
}