Hits:

Indexed by:会议论文

Date of Publication:2019-06-28

Included Journals:EI

Volume:11831 LNAI

Page Number:371-380

Abstract:Biomedical named entity recognition and normalization aim at recognizing biomedical entity mentions from text and mapping them to their unique database entity identifiers (IDs), which are the primary task of biomedical text mining. However, name variation and entity ambiguity problems make this task challenging. In this paper, we leverage domain knowledge by a novel knowledge feature representation method to recognize more entity variants, and model important local context through a dual attention mechanism and a gating mechanism to perform entity normalization. Experimental results on the BioCreative VI Bio-ID corpus show that our proposed system achieves the new state-of-the-art performance (0.844 F1-score for protein/gene entity recognition and 0.408 F1-score for normalization). ? Springer Nature Switzerland AG 2020.