Expert Recommendation Via Semantic Social Networks

The use of social network analysis (SNA) in the design of expert recommendation systems is becoming increasingly popular. However, the experts recommended from such systems often do not meet users’ needs since the network semantic information is largely ignored. In this study, we used conditional logistic analysis to quantitatively examine the semantics of two social networks in a large open source community called Ohloh. It was found that homophily in nationality, location, programming language preference, and community reputation are determinants for forming evaluation and collaboration relationships among the Ohloh members. Moreover, past collaborations and mutual acquaintances are also found to significantly affect the formation of evaluation links but not collaboration links. In addition, we demonstrated how to embed the discovered network semantics into the design of expert recommendation systems through two mechanisms user-based link prediction and Top-N most recognized mechanism.