Identifying advisor-advisee relationships from co-author networks via a novel deep model

Advisor-advisee is one of the most important relationships in research publication networks. Identifying it can benefit many interesting applications, such as double-blind peer review, academic circle mining, and scientific community analysis. However, the advisor-advisee relationships are often hidden in research publication network and vary over time, thus are difficult to detect. In this paper, we present a time-aware Advisor-advisee Relationship Mining Model (tARMM) to better identify such relationships. It is a deep model equipped with improved Refresh Gate Recurrent Units (RGRU). Extensive experiments over real-world DBLP data have well verified the effectiveness of our proposed model.