A Hybrid Method to Extract Triggers in Biomedical Events

Biomedical event extraction from literature is a new research topic in the field of biomedical text mining. Trigger detection is a crucial and challenging subtask in this complex task. A method based on the conditional random field (CRF) model combined with support vector machine (SVM) is developed in this study to detect biomedical event triggers. Rich features, including lexical and structural, are used in model training. Experiment results show the promising perfor-mance of the method, which achieved an F-score of 72.07 on the BioNLP2013-ST corpus. Different from the other similar systems, the valid proteins that participate in the event are identified which is helpful to detect the triggers. We believe this work is a positive contribution to the biomedical text mining community by easing biomedical event extraction. Subject Categories and Descriptors: I.2.7 [Artificial Intelligence]: Biomedical Natural Language Processing Text Mining; 1.5.4 [Pattern Recognition]: Applications – Information Extraction General Terms: Algorithm, Performance