Chemical-disease Relations Extraction Based on The Shortest Dependency Path Tree

Identifying chemical-disease relations (CDR) from biomedical literature could improve chemical safety and toxicity studies. This paper proposes a Shortest Dependency Path Tree (SDPT) to capture the most direct syntactic and semantic relationship between chemical and disease. Based on SDPT, structured dependency features (SDF), structured phrase features (SPF) and flattened dependency features (FDF) are proposed to represent syntactic information between two entities, which are all effective for CDR. Experiments on the CDR training and developing dataset show that our method achieves 55.05% F1-score.