TagLDA based User Persona Model to Identify Topical Experts for Newly Posted Questions in Community Question Answering Sites

摘要

Community Question Answering (CQA) sites are meant for hosting information where user interacts by sharing their knowledge in question-answer form. The user can post their query in the form of questions and wait for other users' to answer. Although the CQA sites provide opportunities for users' to seek help and provide answers in free form text. The huge and semi-structured contents of these sites are posing several challenges. The common challenge is to route the posted questions to the right users' who are experts in the domain of the question resulting faster response. In this paper, we propose a model to route a newly posted question automatically to the right group of experts resulting shorter waiting time. Most of the existing work models the user expertise based on textual contents of Q&A posts. We analyzed the associated tags and voting information additionally to model user expertise for routing the questions. The tag based user persona model (UPM) is presented by leveraging latent dirichlet allocation (LDA) topic modelling on questions answered by the user under tags. The user (answerer) personas generated under question tags are used to find the posterior probability of the user with new questions. Then the user expertise is calculated by multiplying the posterior probability with votes received by the user under question tags. The correctness of model is verified using StackOverflow dataset and the results are compared with existing baseline models.