UTH-CCB@BioCreative V CDR Task: Identifying Chemical-induced Disease Relations in Biomedical Text

This paper describes the system developed by the UTH-CCB team from the University of Texas Health Science Center at Houston (UTHealth), for the 2015 BioCreative V shared tasks of Track 3 on extraction of chemical disease relation (CDR). We participated in both tasks: Task A for “Disease Named Entity Recognition and Normalization (DNER)” and Task B for “Chemical-induced Diseases Relation Extraction (CID)”. For Task A, we developed a Conditional Random Fields based named entity recognition system and used a general Vector Space Model-based approach for entity normalization. To extract the chemical-induced disease relation, we combined two Support Vector Machines-based classifiers, which were trained on sentenceand documentlevel, respectively. Our system achieved a F1 score of 83.53 for Task A and 57.03 for Task B, demonstrating the effectiveness of machine learning-based approaches for automatic extraction of entities and their relations from biomedical literature.