Research On Expert And Answer Recommendation In Online Knowledge Communities

With the rapid development of the Internet,online knowledge communities are emerging in large numbers.As a virtual network platform,the online knowledge community provides users with knowledge sharing services.Registered users in the community can realize knowledge seeking or knowledge sharing by submitting questions or answers without being restricted by time or place.Therefore,conducting knowledge questions and answers through online knowledge communities has gradually become an important means for people to acquire and share knowledge.However,there are a large number of unanswered questions in the community,which directly affect the development of the community and user experience.In addition,due to the differences in user expertise,the quality of the questions or answers submitted is uneven,making users unable to effectively obtain valuable information.Therefore,online knowledge communities have the problem that users cannot obtain high-quality information in a timely manner.In response to the above problems,this article considers improving the efficiency of user acquisition of knowledge from two perspectives.One is to find right community experts to submit complete and high-quality answers to the asker.The other is to find archived answers similar to the new question in the database so as to promptly show the existing solutions to the asker.An expert finding method based on user topic similarity and user professionalism is proposed in order to find expert users who are similar to the asker's professional field.Community users have their own domain preferences and knowledge expertise,and experts with the same or similar preferences are more likely to help the asker,by measuring the topic similarity between the candidate user and the new question,the appropriate expert is selected.At the same time,in order to enable the asker to obtain a high-quality answer,we evaluate the professional level of users based on the historical question and answer data to ensure that the recommended users have certain professional capabilities.After that,the user's authority score in the community is evaluated by comprehensively considering the user's professional field and professional level,and the candidate user's recommendation score is calculated by integrating the user's authority score and the topic similarity between the user and the new question.Finally,an experiment was carried out based on the dataset of the Stack Overflow community,and compared with a variety of existing expert discovery methods.The experimental results show that the expert recommendation method proposed in this paper is generally better than several comparison methods.In order to find archived answers that can better match new questions,we propose an answer recommendation algorithm based on similarity ranking.The algorithm evaluates the answer to be recommended based on the similarity between the new question and the archived answer and the relevance between the new question and the archived answer.First,find multiple similar archived questions from the database based on the similarity between the new question and the archived question,then,the similarity between similar archived questions and archived answers is taken as a consideration factor in evaluating the object to be recommended to assist in selecting appropriate archived answers.Secondly,considering that each answer in the archive may be a similar answer to the new question,the similarity of the archived answer to the new question and similar questions was further evaluated and ranked.After that,the weighted average score is calculated for each archived answer through the weighted average method,and the recommended score of the archived answer is obtained by ranking again.Finally,simulations were performed on the real community dataset,and the simulation results of the proposed algorithm under the indicators of the mean average precision,precision,average reciprocal ranking and normalized discounted cumulative gain are analyzed.The results of various indicators show that the proposed algorithm performs better than the baseline method,and can effectively find appropriate archived answers for new questions.