An empirical investigation of expertise matching within academia
Many organizations have realized that effective management of their knowledge assets is important to survival in today s competitive business environment. Consequently an Organizational Memory (OM) is used to store what has been learned from the past in order that it can be reused by current and future employees. Information retrieval techniques have been widely used to facilitate the retrieval of the right information in an OM at the right time. However, access to information alone is not sufficient since not all knowledge can be transferred into explicit documentation. Expertise, as one of the most important knowledge assets, is normally stored in people s heads and is difficult to codify. Expertise is shared when people communicate with each other. Therefore, finding the right person with the right expertise is recognized as being at least as important as retrieving documents. The typical approaches to find experts include knowledge brokers and expertise database. However, the former approach is impractical in large organizations and geographically disparate organizations whilst the latter approach relies heavily on individuals to specify their expertise and keep updated. This thesis focuses on two questions: (1) How to integrate multiple expertise indications existing in an organizational memory as complementary to the description by experts? (2) How to insure the relevant experts are not overlooked as well as irrelevant experts are minimized? To solve these problems, a conceptual model has been developed so that multiple expertise indications existing in the organizational memory can be semantically integrated. The heterogeneous data sources are integrated by using RDF(S) since RDF allows for a uniform representation of data and RDF Schema represents the conceptual model. In addition, the expertise profiles are extended to include both keyword form and concept form based on the domain ontology; this combined profile integrates the advantages of both keyword search and concept search. A prototype system, which aims to help PhD applicants locate their potential supervisors, has been designed and implemented to test the techniques and ideas. The results of the experiments using real data at the University of Leeds demonstrate the improved performance of expertise matching and also show the advantages of applying semantic web technologies (such as RDF, RDFS, ontologies) to the expertise matching problem.