Large Language Models as Annotators for Machine Translation Quality Estimation

TL;DR

This paper explores using large language models to generate quality estimation annotations for machine translation, enabling training of a more efficient model that achieves competitive results.

Contribution

It introduces a simplified MQM annotation scheme and a systematic prompt design for GPT-4o, improving MTQE training with LLM-generated annotations.

Findings

Annotations correlate well with human judgments.

Training on LLM-generated annotations yields competitive QE performance.

The approach reduces reliance on costly human annotations.

Abstract

Large Language Models (LLMs) have demonstrated excellent performance on Machine Translation Quality Estimation (MTQE), yet their high inference costs make them impractical for direct application. In this work, we propose applying LLMs to generate MQM-style annotations for training a COMET model: following Fernandes et al. (2023), we reckon that segment-level annotations provide a strong rationale for LLMs and are key to good segment-level QE. We propose a simplified MQM scheme, mostly restricted to top-level categories, to guide LLM selection. We present a systematic approach for the development of a GPT-4o-based prompt, called PPbMQM (Prompt-Pattern-based-MQM). We show that the resulting annotations correlate well with human annotations and that training COMET on them leads to competitive performance on segment-level QE for Chinese-English and English-German.