Improving Text Generation Evaluation with Batch Centering and Tempered Word Mover Distance

TL;DR

This paper introduces two techniques, batch-mean centering and tempered Word Mover Distance, to enhance the accuracy of text generation evaluation metrics using BERT-based representations, achieving better correlation with human judgments.

Contribution

It proposes novel statistical and computational methods to improve similarity metrics for text evaluation, leading to state-of-the-art correlation with human ratings.

Findings

Improved correlation with human judgments across benchmarks.

Enhanced statistical properties of contextualized word representations.

State-of-the-art performance in text evaluation metrics.

Abstract

Recent advances in automatic evaluation metrics for text have shown that deep contextualized word representations, such as those generated by BERT encoders, are helpful for designing metrics that correlate well with human judgements. At the same time, it has been argued that contextualized word representations exhibit sub-optimal statistical properties for encoding the true similarity between words or sentences. In this paper, we present two techniques for improving encoding representations for similarity metrics: a batch-mean centering strategy that improves statistical properties; and a computationally efficient tempered Word Mover Distance, for better fusion of the information in the contextualized word representations. We conduct numerical experiments that demonstrate the robustness of our techniques, reporting results over various BERT-backbone learned metrics and achieving state of the art correlation with human ratings on several benchmarks.