Drug Adverse Reaction Discovery Based on Attention Mechanism and Fusion of Emotional Information

This paper proposes a research method of adverse drug reactions based on attention mechanism and fusion of emotional information, and constructs a neural network model, Attention based Convolutional neural networks and Bi-directional long short-Term Memory (ACB). In order to improve the recognition efficiency of adverse drug reactions and solve the problems of gradient explosion and disappearance, it introduced the attention mechanism and Bi-directional Long Short-Term Memory (BiLSTM) to enhance the reliability of the model, as well as mixed together the emotional information of the users’ medication comments. Compared with the superficial information only relied on users’ medication reviews, this is able to enhance features’ expression way and the accuracy of the adverse drug reaction’s classifications. This experiment dataset was based on the local drugs in Xinjiang. The best performance on a test dataset was with ACB obtaining a precision of 95.12% and a recall of 98.48%, and an F-score of 96.77%. The results showed that the ACB model can significantly improve the recognition and classification performance of adverse drug reactions.