Graph-guided context selection improves document translation

Arxiv CS.CL (NLP/language models)

Proposes G^2C-MT, a graph-guided method for selecting context in document-level machine translation. Outperforms retrieval-based and discourse-aware baselines on multiple DocMT benchmarks.

nlp

Arxiv CS.CL (NLP/language models)·Baijun JiZixuan ZhouXiangyu DuanYu LiuLongbo SunRupu WeiBohong Zhao

8 days ago