Research On Discovering Domain Experts In Online Q&A Communities

Accurately finding expert users with high professional and influence in various fields in the Q&A community will help the system to accurately recommend questions to them and make the questions be answered accurately and professionally.Also,it helps to raise the knowledge level of Q&A community knowledge bases,increases the user's participation and community activity,and provides high quality information resources for external search engines.At present,research on experts finding in online Q&A community generally adopts the method based on topic or the method based on link analysis.The method based on topic model generally describes the users' interest distribution or professional distribution use the generated content by users.The method based on link analysis mainly utilizes the link relationship network among users to calculate the users' influence.This paper consider the advantages of the two methods to research how to combine the users'professionalism and influence to find domain expert users in online Q&A community.Aiming at calculating the professional level of users,this paper is inspired by the interactive process of online Q&A community that an answerer usually choose a question by the question's describe and his or her domain knowledge before answer it.Thus,we propose a Question-Answerer-Topic(QAT)model to modeled the Question-Answerer pair's topic distribution.The model can effectively combine the information of the questions and the answerer to improve the aggregation effect of the topics.Then we based on the question-answerer's topic distribution derived by QAT model to integrate the number of vote in each answer as the weighting factor to evaluate the user's professional level,and then calculate the user's professional level in each topic in the field.Aiming at calculating the users' influence,based on the users' professional level in each topic,we considered combining the link analysis method,and then improved the topic-sensitive PageRank algorithm.The user's topical professional level is taken as the factor affecting the remote jump in the random jump in the improved model.Finally,a domain experts finding method that combines domain topics,votes data,and link analysis is proposed to calculate the final scores of users in different topics in the domain.And the words in each topic can also be used as a matching factor when the community system recommends a new question to domain experts.This paper collected data from the Chinese online Q&A community Zhihu in the field of artificial intelligence as a domain Q&A dataset.Then the data extraction,the relation extraction and the necessary text processing are carried out to form a standardized domain Q&A dataset.Finally,using the standardized dataset,the proposed method of field experts finding was compared and analyzed experimentally.The experimental results show that the domain experts finding method proposed in this paper can effectively find expert users,and its effect is better than the classic experts finding methods in online Q&A community.