Extracting Personal Conceptual Structures from PC Documents and its Application to Web Search Personalization

This paper proposes a method of extracting personal conceptual structures from documents on a personal computer that contain a great deal of personal information, and applying them to personalize Web searches. Everybody has differing ideologies, concepts, and knowledge and there is a lot of personal information stored on PCs. While it is easy for a person to determine what a PC user thinks and knows, computers cannot. In this work, two types of personal conceptual structures are represented, i.e., a "personal concept tree" and "relationships between terms". The "personal concept tree" indicates personal concept classification. Relationships between terms" indicates how the user thinks of a term, and how it is created from the "personal concept tree" based on deviations in the appearance of the term. This paper also proposes the personalization of Web searches, i.e., expanding query keywords and re-ranking of search results using personal conceptual structures.