A LSTM-based method with attention mechanism for adverse drug reaction sentences detection

Abstract

Adverse drug reactions (ADRs) are among the top causes of morbidity, mortality and substantial healthcare costs and thus should be detected early to reduce consequences on health outcomes. Many conventional machine learning based methods have been presented to automatically detect adverse drug effect (ADE) mentions from biomedical texts. However, owing to the complexity of natural language text in the biomedical domain, some ADE mentions might not be detected. In this paper, we propose a Long Short-Term Memory with Attention (LSTMA) which incorporates attention mechanism and LSTM network to address the problem of automatic detection of ADR assertive text segments from biomedical texts. Experimental results on standard ADE dataset show that the proposed method outperforms significantly the state-of-the art methods for ADR class with an F-scores of 0.89.