Data integration in the wild: from instances to concept catalogs
Patrick Pantel, Andrew Philpot, Eduard H. Hovy · 2007
Internet users today have access to seemingly unlimited amounts of information, and yet, due to the lack of companion structure, such information often cannot be effectively located or used. In response, we propose and demonstrate one component of a framework for autonomously learning and classifying instance data drawn from document collections on the Web. We have already applied this general-purpose prototype to extract conceptual representations from various websites for DVD products, hotels, restaurants, people, and cars.