ExpRec: Deep knowledge-awared question routing in software question answering community

Software question answering community (SQAC) as an effective platform of knowledge sharing has achieved rapid development. In SQAC, one critical and challenging problem is question routing (or expert recommendation). To solve this problem, previous approaches focus on learning the relevance between the question and answerers. However, such approaches usually suffer from the data sparsity and noise issues which may reduce the accuracy of the question routing. Moreover, previous approaches also ignored the response quality and timeliness of the question routing. To tackle those issues, we study the question routing problem from two aspects: 1) the answerer’s relevance to the given question, and 2) the answerer’s capability. We first propose a deep knowledge-awared question routing framework (termed ExpRec) which leverages the attentive embedding propagates and their high-order connectivities to learn the answerer’s relevance to the given question. Then we explicitly model the answerer’s capability and incorporate it with the answerer’s relevance to the given question. Finally, to evaluate the performance of ExpRec, we conduct extensive experiments on two real-world datasets. The experimental results show that ExpRec outperforms other five state-of-the-art approaches significantly.