Research On Recognition And Classification Of Adverse Drug Reactions Based On Deep Learning

Adverse Drug Reactions(ADR)refer to a series of harmful reactions brought by drug quality issues or nonstandard use of drugs to different extents.Common adverse drug reactions include side effects,residual effects(post-effects)and idiosyncratic reactions,etc.At present,adverse drug reactions are becoming the focus of medical R&D(research and development)departments,common users and major medical institutions.So the safety of user’s medication is especially highlighted.Considering the bottleneck of researches into new drugs in medical research and development platforms,it’s impossible to research and develop drugs as well as conduct drug testing for a large number of people in a limited time.Therefore,researching on ADR and grasping known and potential drugs’ adverse reactions are of great value to medical research and development.Researching on the adverse drug reactions of Xinjiang local drugs can provide theoretical guidance for a better research and development of clinical drugs.At present,the research on adverse drug reactions is mainly based on English corpus.There is a lack of deep learning algorithm in research technology,and there are few researches focusing on the use of corpus context information and semantic information.Therefore,in view of the above problems and based on deep learning technology,this paper carried out a research on Xinjiang local entity recognition and classification of adverse drug reactions the following three aspects:(1)The corpus of adverse drug reactions of Xinjiang local were constructed.At present,there is a lack of public data sets related to adverse drug reactions at home and abroad.In view of this problem,this paper uses the crawler technology based on Python language to crawl the text information of user comments on Baidu Post Bar,Sina Weibo,Zhihu and other major user drug consultation websites,and carries out data cleaning,word segmentation and labeling work to provide data support for the later study of adverse drug reactions in Xinjiang.(2)Xinjiang local entity recognition of adverse drug reactions was conducted.For deep learning model’s inadequate capability of learning a small amount of data,an entity recognition method for adverse drug reactions based on transfer learning is proposed to construct a joint neural network model SAIBC.The transfer learning makes full use of Chinese word segmentation and Xinjiang local adverse drug reactions entity recognition to obtain the shared boundary information.At the same time,the specific information of the Chinese word segmentation is filtered.Introduction of Independently Recurrent Neural Network(Ind RNN)can improve the feature’s expression ability and enable SAIBC models to focus on important information of different entities from different levels,thus better capturing the dependencies of long sentence.(3)A researching approach to adverse drug reactions classification based on Attention mechanism and combined with emotion information was proposed.Then,a neural networks ACB model jointed by Attention mechanism,Convolutional neural networks and Bi-directional Long Short-Term Memory was constructed.To make the recognition more accurate and handle the gradient exploding and gradient vanishing issue,Attention mechanism is introduced to enhance the model’s reliability.Meanwhile,compared with the superficial information only dependent on users’ comments on medication,the emotion information merging with users’ comments on medication in this paper can further strengthen the feature’s expression ability and accuracy of adverse drug reactions classification.In this task,ACB model performs well with the accuracy,recall rate and F1-Score being 95.12%,98.48% and 96.77%,respectively.