Subgraph Representation Learning for Team Mining

BSTRACT Team mining is concerned with the identification of a group of experts that are able to collaborate with each other in order to collectively cover a set of required skills. This problem has mainly been addressed either through graph search, which looks for subgraphs that satisfy the skill requirements or through neural architectures that learn a mapping from the skill space to the expert space. An exact graph-based solution to this problem is intractable and its heuristic variants are only able to identify sub-optimal solutions. On the other hand, neural architecture-based solutions are prone to overfitting and simplistically reduce the problem of team formation to one of expert ranking. Our work in this paper proposes an unsupervised heterogeneous skip-gram-based subgraph mining approach that can learn representations for subgraphs in a collaboration network. Unlike previous work, the subgraph representations allow our method to mine teams that have past collaborative history and collectively cover the requested desirable skills. Through our experiments, we demonstrate that our proposed approach is able to outperform a host of state-of-the-art team mining techniques from both quantitative and qualitative perspectives.