Creating Models of Real-World Communities with ReferralWeb
Henry Kautz · 1998
online community. In general, however, you want to find experts who are members of existing "real world" communities. Furthermore, the best, busiest experts are probably the least likely ones to bother to register with any kind of "expert locator" service. Therefore it is necessary to generate social networks without the explicit cooperation of any of the people who appear in the networks. ReferralWeb does this by mining information from publicly available documents. We are experimenting with different ways to determine the "links" between people using public documents, using different kind of (off line) spiders to populate social networks. ffl In scientific communities, the "co-author" relationship is a reliable indicator of connectedness. One of our spiders downloads and parses bibliographic databases. Our current demonstration system contains a network of about 10,000 computer science researchers in the areas of AI, natural language processing, and theory. ffl