The long-standing one-to-many issue of the open-domain dialogues poses significant challenges for automatic evaluation methods, i.e., there may be multiple suitable responses which differ in semantics for a given conversational context. To tackle this challenge, we propose a novel learning-based automatic evaluation metric (CMN), which can robustly evaluate open-domain dialogues by augmenting Conditional Variational Autoencoders (CVAEs) with a Next Sentence Prediction (NSP) objective and employing Mutual Information (MI) to model the semantic similarity of text in the latent space. Experimental results on two open-domain dialogue datasets demonstrate the superiority of our method compared with a wide range of baselines, especially in handling responses which are distant to the golden reference responses in semantics.
@article{arxiv.2305.16967,
title = {Evaluating Open-Domain Dialogues in Latent Space with Next Sentence Prediction and Mutual Information},
author = {Kun Zhao and Bohao Yang and Chenghua Lin and Wenge Rong and Aline Villavicencio and Xiaohui Cui},
journal= {arXiv preprint arXiv:2305.16967},
year = {2023}
}