The Attention Based BLSTM Model Integrating Sentence Embeddings for Biomedical Event Trigger Identification

Abstract

Biomedical event extraction is an important and challenging task in Information Extraction, which plays an important role for medicine research and disease prevention. Trigger identification has attracted much attention as the prerequisite step in biomedical event extraction. To skip the manual complex feature extraction, we propose a trigger identification method based on Bidirectional Long Short Term Memory (BLSTM) neural network. To obtain more semantic and syntactic information, we train dependency-based word embeddings to represent words, and add sentence embeddings to enrich sentence-level features. In addition, the attention mechanism is integrated to capture the most important semantic information in the sentence. The experimental results on the multi-level event extraction (MLEE) corpus show that the proposed method outperforms the state-of-the-art systems, achieving an F-score of 79.96%.