Evaluating Text Style Transfer Evaluation: Are There Any Reliable Metrics?

TL;DR

This paper assesses the reliability of various automatic metrics, including large language models, for evaluating text style transfer tasks across multiple languages, aiming to find more effective evaluation methods.

Contribution

It systematically evaluates existing and novel metrics, including LLM-based approaches, for TST evaluation and demonstrates their effectiveness through correlation with human judgments.

Findings

Advanced NLP metrics improve evaluation accuracy.

LLM-based evaluations outperform traditional metrics.

Ensemble approaches enhance reliability of TST assessment.

Abstract

Text style transfer (TST) is the task of transforming a text to reflect a particular style while preserving its original content. Evaluating TST outputs is a multidimensional challenge, requiring the assessment of style transfer accuracy, content preservation, and naturalness. Using human evaluation is ideal but costly, as is common in other natural language processing (NLP) tasks, however, automatic metrics for TST have not received as much attention as metrics for, e.g., machine translation or summarization. In this paper, we examine both set of existing and novel metrics from broader NLP tasks for TST evaluation, focusing on two popular subtasks, sentiment transfer and detoxification, in a multilingual context comprising English, Hindi, and Bengali. By conducting meta-evaluation through correlation with human judgments, we demonstrate the effectiveness of these metrics when used individually and in ensembles. Additionally, we investigate the potential of large language models (LLMs) as tools for TST evaluation. Our findings highlight newly applied advanced NLP metrics and LLM-based evaluations provide better insights than existing TST metrics. Our oracle ensemble approaches show even more potential.