Suggesting Questions that Match Each User’s Expertise in Community Question and Answering Services
Katsunori Fukui, Tomoki Miyazaki, Masao Ohira · 2019
Stack Overflow has been already recognized as an indispensable CQA (Community Question and Answering) service for developers. However developers as questioners sometimes cannot get any answers or they only can get incomplete answers to resolve their questions, since over several thousands of questions and answers are posted on a daily basis and expert developers are currently facing with a difficulty in finding questions which can be better or best answered by them. In order address the issue, our study aims at developing a bot that helps expert developers find questions matching each developer's expertise. The existing study [8] proposed a promising approach to the expert recommendation for CQA services, based on PMF (Probabilistic Matrix Factorization) algorithm. In order to improve the accuracy of the expert recommendation, in this paper we present our approach which combines the PMF (Probabilistic Matrix Factorization) approach and a term expansion techniques using word embeddings. As a result of an experiment comparing our approach with the existing approach only based on PMF, we show that our approach based on PMF and the term expansion with Word2vec and fastText slightly outperforms the existing approach only based on PMF.