MixRED Dataset

To facilitate the MixRE task, we introduce the first human-annotated Mix-lingual Relation Extraction Dataset (MixRED). MixRED is meticulously constructed by blending English, Chinese, Japanese, and German versions of documents through a systematic construction framework. This framework identifies crucial elements in texts and emphasizes them to generate mix-lingual content, aiming to optimize the potential impact on model performance.