MetaHOPE: A Metaphor-Oriented Evaluation Framework for Analysing MT and LLM Translation Errors

This opinion paper proposes MetaHOPE, an error severity-aware annotation framework designed to evaluate metaphor translations in machine translation (MT) and large language models (LLMs). The authors address the semantic complexity and cultural embeddings that often cause ambiguity for NLP models.

• The study evaluates GoogleMT, GPT5.4, and Hunyuan-7b as representative Neural MT and LLM systems.
• It utilizes two human-annotated metaphor corpora, VUAMC and PSUCMC, for English-to-Chinese and Chinese-to-English translation tasks.
• The researchers produced new bilingual resources by applying the MetaHOPE framework to monolingual source corpora and generating human post-edited gold references.

The authors believe this evaluation framework, along with the shared parallel corpora resources and error analysis, can provide useful insights for the field of metaphor translation study.