REAM$\sharp$: An Enhancement Approach to Reference-based Evaluation Metrics for Open-domain Dialog Generation

TL;DR

This paper introduces REAM#, a method to enhance reference-based evaluation metrics for open-domain dialogue systems by predicting and augmenting reference sets to improve their reliability and correlation with human judgments.

Contribution

It proposes a prediction model to assess and improve the quality of reference sets, thereby increasing the reliability of reference-based evaluation metrics.

Findings

The prediction model effectively estimates reference set reliability.

Augmenting reference sets with predicted high-quality references improves metric reliability.

Enhanced metrics show better correlation with human judgments.

Abstract

The lack of reliable automatic evaluation metrics is a major impediment to the development of open-domain dialogue systems. Various reference-based metrics have been proposed to calculate a score between a predicted response and a small set of references. However, these metrics show unsatisfactory correlations with human judgments. For a reference-based metric, its reliability mainly depends on two factors: its ability to measure the similarity between the predicted response and the reference response, as well as the reliability of the given reference set. Yet, there are few discussions on the latter. Our work attempts to fill this vacancy. We first clarify an assumption on reference-based metrics that, if more high-quality references are added into the reference set, the reliability of the metric will increase. Next, we present REAM$\sharp$: an enhancement approach to Reference-based EvAluation Metrics for open-domain dialogue systems. A prediction model is designed to estimate the reliability of the given reference set. We show how its predicted results can be helpful to augment the reference set, and thus improve the reliability of the metric. Experiments validate both the effectiveness of our prediction model and that the reliability of reference-based metrics improves with the augmented reference sets.