Research On Drug Relationship Mining For Social Media

Adverse Drug Reactions(ADR)refers to the physiological effects and even physicaldamage caused by patients after using drugs.At present,more and more patients will have a series of adverse drug reactions caused by taking drugs,including those caused by taking one or more drugs,and the adverse reactions caused by the introduction of drugs on the market every year in the world.The number of hospitalizations is as high as one million.Due to the complexity of the mechanism of drug action,the complexity of drugs and drugs,and the differences in individual constitution between patients,it is difficult to fully investigate all adverse reactions of drugs in medical literature reports.Therefore,in order to solve this problem,it is necessary to quickly and effectively capture previously undiscovered side effects,expand to use a wider range of data sources to detect ADR,and then tap more drug relationships to promote the development of biomedicine and protect the lives of patients health and safety.With the rapid development of the Internet,Twitter,Reddit,Weibo and other social media have become the main platforms for the public to exchange emotions and share knowledge,and can also be used for trend tracking and real-time information retrieval,etc.,which is also a drug in the field of biomedicine.Relationship mining research has brought richer sources of information and data.But at the same time,it also brings a lot of challenges.For example,there are millions of tweets and many informal expressions in social media every day on social forums.Therefore,how to effectively identify ADR-containing ones from a large number of social posts Posts and how to dig out the relationship between drugs and adverse reactions in the article are urgently needed to be solved in biomedicine.This paper is based on social media for drug relationship mining research.Its realization process is mainly divided into four parts: social media data acquisition,named entity identification of drugs and ADR,relationship extraction between drugs,and drug adverse reaction post detection.In this paper,the experimental data focuses on social networks(Twitter)and social forums(related to breast cancer),using crawler technology to crawl its posts and perform text preprocessing during the data acquisition stage.In the entity recognition stage of drugs,symptoms and ADR,based on the BILSTM + CRF model,incorporating the pre-trained language model Bio-BERT based on the medical literature as a word representation,and adding the Self-Attention mechanism to better handle long-distance dependence Problem while learning the internal structure of the sentence.In the task of extracting the relationship between drugs and ADR,the Glove word vector predicted by social media Twitter is used to stitch with the position features of the entity,and the BILSTM + Muti Head-Attention model is used to mine the relationship between drugs.In the detection of adverse drug reaction posts,BILSTM and Capsule networks were constructed as experimental models,and dictionary matching technology and character-level vector features were added to solve the informal expression of language in social networks,thereby improving the accuracy of classification.This article provides a reference value for further research on the relationship between drugs,which can reduce the time and money consumption in the ADR discovery process.