BioRel: A Large-Scale Dataset for Biomedical Relation Extraction

Valuable biomedical knowledge usually exists in the form of electronic publications and literature, which is growing at an enormous rate. Relation extraction plays a critical role in discovering such knowledge and transform them into structural form. Previous relation extraction datasets in biomedical domain are mainly human-annotated, whose scales are usually limited due to their labor-intensive and time-consuming nature. In this paper, we present BioRel, a large-scale dataset constructed by using Unified Medical Language System (UMLS) as knowledge base and Medline as corpus. Entities in sentences of Medline are identified and linked to UMLS by Metamap. Relation label for each sentence is recognized using distant supervision. We adapt both state-of-the-art deep learning and statistical machine learning methods as baseline models and conduct comprehensive experiments on BioRel. Experimental results show that BioRel is suitable for training and evaluating relation extraction models for both deep learning and statistical methods by providing both reasonable baseline performance and many remaining challenges.