Structures or Texts? A Dynamic Gating Method for Expert Finding in CQA Services

Zhiqiang Liu, Yan Zhang

Abstract: Expert finding plays an important role in community question answering websites. Previously, most works focused on assessing the user expertise scores mainly from their past question-answering semantic features. In this work, we propose a gating mechanism to dynamically combine structural and textual representations based on past question-answering behaviors. We also use some user activities including temporal behaviors as the features, which determine the gate values. We evaluate the performance of our method on the well-known question answering sites Stackexchange and Quora. Experiments show that our approach can improve the performance on expert finding tasks.