Document-level Relation Extraction

Relation extraction (RE) has been widely studied by previous work as an important aspect in the information extraction (IE) process, and recently there is a growing interest to investigate document-level RE as a more realistic setting. In this presentation, I will review the recent development of document-level RE and multiple supported datasets that are introduced within the last two years. Specifically, we are interested in the potentials of joint inference of both RE and coreference resolution for document understanding, making use of multiple existing resources/datasets of sentence/document-level RE and coreference resolution.