PAIR: Person-Atrribute Identification and Retrieval

The Web has become the most major information source of our daily activities. In this project, we consider a special information need: finding the instance-attribute information from the Web, and particularly, we focus on the instance domain of person. Examples include the contact information (attribute) and research interests (attribute) of faculty or students (person). Most of such personal information already exists today as conventional, unstructured pages because people find it easier to create them that way. Such type of information need is very common but cannot be directly supported by current keyword-matching-based search engines. People get used to using a two-phase search scenario: locating the candidate pages first and then getting into them for the desired information piece. Based on the stimulation of such human search behavior, we design a blockbased retrieval engine, upon the general search engines, to help the finding of person-attribute information from the Web. The experiment on several faculty members has shown the feasibility of the approach. It is believed that the approach can benefit not only the finding of such information by users, but also various extraction applications.