Temporal expert profiling: With an application to T-shaped expert finding

Abstract

Cost optimization continues to be a critical concern for many human resources departments. The key is to balance between costs and business value. In particular, computer science organizations prefer to hire people who are expert in only one skill area and have a slight superficial knowledge in other areas that gives them the ability to collaborate across different aspects of project. Community Question Answering networks provide good platforms for people and organizations to share knowledge and find experts. An important issue in expert finding is that an expert has to constantly update his knowledge after being saturated in his field of expertise to still be identified as expert. A person who fails to preserve his expertise is likely to lose his expertise. This work justifies this question that does take the concept of time into account improve the quality of expertise retrieval. We propose a new method for T-shaped expert finding that is based on temporal expert profiling. The proposed method takes the temporal property of expertise into account to mine the shape of expertise for each candidate expert based on his profile. To this end, for each candidate expert, we take snapshots of his expertise trees at regular time intervals and learn the relation between temporal changes in different expertise trees and candidates’ profile. Finally, we use a filtering technique that is applied on top of the profiling method, to find shape of expertise for candidate experts. Experimental results on a large test collection show the superiority of the proposed method in terms of quality of results in comparison with state-of-the-art.