Question Routing by Modeling User Expertise and Activity in cQA services

抄録

Community Question Answering (cQA) sites such as Yahoo! Answers and Stack Overflow have emerged as a new type of community portals that allows users to answer the questions asked by other people. The cQA archives have accumulated a huge mass of questions and answers. On account of the progressively increasing questions, there are many questions can be solved or answered by others efficiently. In this paper, we address the problem of efficient question routing. We present a new approach that combines user's expertise and user's activity to solve this problem. First, we model user's expertise by the contents of user's answering questions in the past, and we enhance user's expertise by social network characteristics in a cQA portal. Second, we model and predict user's activity by analyzing the distribution of their previous answering records. Experiments conducted on a real cQA data, Stack Overflow, show that our approach leads to a better performance than other baseline approaches significantly. In terms of the evaluation metric, MRR, the performance of the content-based approach is 0.0999 and that of our method is 0.1372 respectively. We can get a 37.34% improvement over the traditional content-based method. On average, each of 6,160 test questions gets at least one answer if it is routed to the top 7 ranked users by our approach.