Answer Quality Prediction Based On The Influence Of User's Behavior

With the rapid development of the Internet,the traditional search engine based on keyword matching can't understand the user's intention very well.The accuracy and recall of the information retrieval system can't meet the demand of information retrieval.Under this background,community question answering system has become one of the hotspots in the field of artificial intelligence.The questions and answers in the community question answering system are given by the user in natural language,which not only meets the needs of the public to acquire and share knowledge well,but also understands the intention of the user well.The emergence of community question answering system to overcome the shortcomings of the search engine.With the growing number of community question answering system,the quality of answers provided by different users is enormous,and the quality of the answers returned to the site is reduced.Therefore,it has great significance to identify and filter low-quality answers and select high-quality answers,which will improve the speed of retrieval and the quality of community question answering system.In this paper,an answer quality evaluation model is proposed by combining the existing methods of answer quality evaluation.The model combines the user's personal information and the user's social attributes to assess the quality of the answer.Where the personal information of the user includes the attribute information of the answer,the answer behavior characteristic and the answer time characteristic.First of all,this paper extracts the user characteristics of the answers from three aspects of the user's personal information,it is found that the time characteristics of the respondents have an extremely important effect on the quality of the answers.Secondly,LDA algorithm is used to get the topic relevance among users,and the topic similarity between users is introduced into the PageRank algorithm to obtain the user-related authoritative value.In the end,this paper uses SVM model to combine the personal information of the users,the optimized PageRank,the vote number of users to evaluate the quality of the answers comprehensively.We take the Stack Overflow test set to evaluate our proposed model,the experimental results show that our proposed model is higher than the existing evaluation methods.