Designing Precise and Robust Dialogue Response Evaluators
This work addresses the need for more reliable and robust automatic evaluators in dialogue systems, offering an incremental improvement over existing methods.
The authors tackled the problem of automatic dialogue response evaluation by proposing a reference-free evaluator using semi-supervised training and pretrained language models, achieving a strong correlation (>0.6) with human judgement and robust generalization.
Automatic dialogue response evaluator has been proposed as an alternative to automated metrics and human evaluation. However, existing automatic evaluators achieve only moderate correlation with human judgement and they are not robust. In this work, we propose to build a reference-free evaluator and exploit the power of semi-supervised training and pretrained (masked) language models. Experimental results demonstrate that the proposed evaluator achieves a strong correlation (> 0.6) with human judgement and generalizes robustly to diverse responses and corpora. We open-source the code and data in https://github.com/ZHAOTING/dialog-processing.