Document-level relation extraction (RE) task aims to predict predefined relation types of
every entity pair in a given document. Compared with the sentence-level counterpart …