Research On Question Classification And Expert Discovering In Online Q&A Community

The advent of Web3.0 and the era of big data has provided more interactive ways for network users,and a series of community platforms for user interaction have also emerged.Network Q&A communities are one of them.Users can ask questions and answer other users' questions in the community platform.Zhihu,as the Chinese version of Quora,has been receiving attention from a large number of Internet users since its registration opened in 2013.Currently,Zhihu has developed into China's largest online Q&A community.The online Q&A community is like a giant knowledge base,providing users with a platform for online communication and experience sharing.A large amount of user-generated content in the online Q&A community has great value.How to discovery the user's interest preferences,stimulate the enthusiasm of more users,and improve the quality of user-published content is crucial to the spread of knowledge in the community and the improvement of community's platform influence.In this paper,by constructing question classification model and expert discovering model,the automatic classification of questions and the discovering of domain experts are realized,so that the questions raised by users can be quickly categorized,and then experts in the field are found,and the questions are pushed to the experts to answer.The mainly work completed is as follows:(1)A multi-channel question feature extraction model is constructed.When implementing automatic classification of community question texts,the LDA topic model and Doc2 vec model were used to extract the question text features,and a multi-channel question feature extraction model was constructed.This method effectively solved the problem of short text content and sparse data.(2)An integrated classification method for questions in online Q&A community is proposed.The multi-channel fused question text features are input to a support vector machine and a random forest to obtain a base classifier,and the base classifier is integrated through the bagging method to obtain the final classification result.(3)Completed the construction and implementation of the online Q&A community expert discovering model.The topic distribution of user responses is extracted through the LDA model,and the similarity of user expertise is calculated.The number of users' followers,the number of answers,and the number of likes are taken as the user's own influence,and the user-to-user attention relationship and expertise similarity are used toconstruct a user-to-user probability transfer matrix.We can calculate the professional ranking of users,and get highly active and influential expert users.