Research On Knowledge Mining Of Relationship Between Food And Diseases In Online Health Q&A Community

With the improvement of people's awareness of health management,patients are paying attention to active self-health management.The online health Q&A community provides a new way for communication for doctors and patients and provides access to basic health medical knowledge.The online health community includes a large number of questions and answers.These questions and answers data have the characteristics of social media data: rapid update,false information,and closeness to the patient's daily health management needs.Therefore,knowledge mining based on online health community has complementary content to mainstream literature knowledge,and can discover the latest health medical knowledge.In this article,we use the Q&A data from XYWY(a Chinese health website)to find out how to mine healthcare related knowledge from the online health community,which mainly includes automatic extraction of knowledge and reliability evaluation of knowledge quality.First of all,the relationship food between diseases is taken as an example to study the automated extraction of knowledge in online health communities.Using multi-source dictionary matching for entity name recognition and convolutional neural network method for relationship classification,the relationship between common 30 diseases and 215 food ingredients was extracted.The model F1(comprehensive index of both accuracy and recall)is 0.80,while the accuracy of the manual annotation is 0.84 and the F1 of the support vector machine is 0.74.On this basis,this paper evaluates the reliability of the extracted knowledge.In terms of assessment methods,the paper constructs text features,time features,question and answer relevance features,respondents' personal information features,search engine-driven features,topical features,and machine learning methods for automated evaluation of answers.The empirical results show that these features can significantly improve the performance of the model.The gradient boosting decision tree model as a whole is superior to the support vector machine and the logistic regression model.This article has two innovations.The first is to enrich the text mining research of online health community and expand the research face of health management knowledge mining.This article applied the deep learning method to online health community knowledge mining to realize the automatic mining of medical effect of food.Second,current knowledge mining does not take into account the quality of knowledge mining.Based on knowledge mining,the article focuses on the reliability assessment model of knowledge and provides new ideas and methods for solving the knowledge contradiction in online communities.