Capturing scholar's knowledge from heterogeneous resources for profiling in recommender systems

Abstract

In scholars' recommender systems, acquisition knowledge for construction profiles is crucial because profiles provide fundamental information for accurate recommendation. Despite the availability of various knowledge resources, identification and collecting extensive knowledge in an unobtrusive manner is not straightforward. In order to capture scholars' knowledge, some questions must be answered: what knowledge resource is appropriate for profiling, how knowledge items can be unobtrusively captured, and how heterogeneity among different knowledge resources should be resolved. To address these issues, we first model the scholars' academic behavior and extract different knowledge items, diffused over the Web including mediated profiles in digital libraries, and then integrate those heterogeneous knowledge items by Wikipedia. Additionally, we analyze the correlation between knowledge items and partition the scholars' research areas for multi-disciplinary profiling. Compared to the state-of-the-art, the result of empirical evaluation shows the efficiency of our approach in terms of completeness and accuracy.