Quality Assessment of Chinese-French Translation Based on Language Model

This study aims to leverage the human language understanding capabilities of large language models in the field of translation evaluation. By employing methods such as fine-tuning, chain-of-thought (CoT), and prompt template techniques for large language models, combined with classic translation evaluation approaches, we construct a novel automated system for assessing the quality of Chinese-French translation outputs. The research is divided into two main aspects: first, an investigation into quantifiable Chinese-French translation quality evaluation methods based on BERT pre-trained language model; second, an exploration of generative language model-based evaluation approaches using the Qwen2.5 model that can produce specific scoring rationales for Chinese-French translations.