xCOMET: Transparent Machine Translation Evaluation through Fine-grained Error Detection
This addresses the need for more interpretable evaluation in machine translation, offering a tool that helps researchers and practitioners identify specific errors, though it is incremental as it builds on existing learned metrics.
The paper tackles the lack of transparency in machine translation evaluation metrics by introducing xCOMET, which integrates sentence-level scoring with error span detection, achieving state-of-the-art performance across evaluation types and providing error categorization.
Widely used learned metrics for machine translation evaluation, such as COMET and BLEURT, estimate the quality of a translation hypothesis by providing a single sentence-level score. As such, they offer little insight into translation errors (e.g., what are the errors and what is their severity). On the other hand, generative large language models (LLMs) are amplifying the adoption of more granular strategies to evaluation, attempting to detail and categorize translation errors. In this work, we introduce xCOMET, an open-source learned metric designed to bridge the gap between these approaches. xCOMET integrates both sentence-level evaluation and error span detection capabilities, exhibiting state-of-the-art performance across all types of evaluation (sentence-level, system-level, and error span detection). Moreover, it does so while highlighting and categorizing error spans, thus enriching the quality assessment. We also provide a robustness analysis with stress tests, and show that xCOMET is largely capable of identifying localized critical errors and hallucinations.