PSAbot: A Chatbot System for the Analysis of Posts on Stack Overflow

Abstract

With the progressive development of technology, programming learners have significantly increased. However, the lack of human tutors and the rapidly updating information cause the learners to spend a considerable amount of time browsing and filtering authentic online resources, and decrease learning efficiency. Although many coding websites and programming communities can provide credible advice, it is still a challenge for learners to figure out their accurate questions. Therefore, we devised a Chatbot system, named PSAbot, to consider the above issue. PSAbot supports keyword extraction and analysis for multiple posts to guide the users through questions. PSAbot applies word embedding, sentence similarity, LDA (Latent Dirichlet Allocation) topic modeling, and weighting functions to help filter out redundant information and decrease the time cost of browsing, and further improve the learning efficiency. The conducted experiments show that about 80% of the Top1 answers recommended by PSAbot can largely meet the user expectations.