Abstract:Biomedical named entity recognition (BNER) is a critical task for biomedical information extraction. Most popular BNER approaches based on deep learning utilize words and characters as features to represent medical text. However, many medical terminologies are composed of multiple words and characters, and semantic ambiguity occurs when splitting them into fragments. So, the standard attention mechanism is challenging to focus on entities by word or character embedding categorical inference. This paper proposes a dictionary-guided attention model for BNER. Initially, we extracted medical concepts as large-size words to supplement comprehensive semantic information of the medical terminology by matching electronic medical record (EMR) text to the medical dictionary; as a result, the concepts and characters are incorporated for feature representation. Later, based on the matched dictionary results, an adaptive attention strategy is proposed to focus on the medical concept to assign higher attention weight to characters contained in a concept. Furthermore, semi-supervised learning is introduced to reduce the manual labeling of data and to handle the entities not defined in the medical dictionary. The proposed model is evaluated using a real-world EMR dataset that induced superior performance compared to existing state-of-the-art methods.

Dictionary Guided Attention Network for Named Entity Recognition in Chinese Emrs