A hybrid text mining system for chemical entity recognition and classification using dictionary look-up and pattern matching @ BeCalm challenge evaluation workshop

Chemicals as therapeutics and investigational agents receive much attention in clinical research and applications recently. However, automated approaches to recognize and categorize the chemical entities in biomedical text are challenging because of the wide varieties of morphologies and nomenclature. We present here a hybrid text mining system that combines chemical lexicon and patterns for recognition/categorization. We applied this approach to identify chemical entities from the patent abstracts of BioCreative V.5 Chemical Entity Mention Recognition (CEMP) corpus. We also FRPSDUHG WKH K\EULG DSSURDFK ZLWK WKH 3WUDGLWLRQDO ́ lexion-based method, and illustrated that the hybrid approach can achieve enhanced performance (i.e. precision, recall, and F-score) than the lexion-based method.