2007 Volume 19 Issue 6 Pages 670-679
This paper describes a new approach to extract affiliations of researchers from the Web. In the context of Semantic Web and information retrieval, there have been many studies on social network mining and utilization. In such systems, it is important to obtain personal metadata. In this paper, a novel algorithm using a search engine and machine learning is proposed to output the affiliation for a given researcher. First, given a researcher name, we query to a search engine with a combination of the name and candidate affiliations. Using the hit counts of a search engine, we measure the strength of co-occurrences, and provide the most possible affiliations. Then, web pages including the researcher's CV or profile are sought in order to confirm the affiliations. We evaluate our system on 60 researchers and show effectiveness of the algorithm as well as the scope and limitations.