Collecting precise knowledge from scholars' context for profiling is crucial in recommender systems as profiles provide foundational information for successful recommendation. However, acquiring of scholars' knowledge is often a challenging task since it is associated with difficulties including: what are the appropriate knowledge resources, how knowledge items can be unobtrusively captured, and how heterogeneity among different knowledge sources should be resolved. Despite the availability of various knowledge resources, identification and collecting comprehensive knowledge in an unobtrusive manner is not straightforward. To address these issues, we analyze the scholar academic behaviors and collect various scholars' knowledge diffused over the Web. The result of empirical evaluation shows the efficiency of our approach in terms of completeness and accuracy.