A Hybrid Framework for a Personalized Document Filtering System

摘要

As more information becomes available electronically, tools for finding Information of interest to users become increasingly important. In this research, a personalized document filtering system with learning ability is designed and implemented. Each user puts one query in the system and five intelligent software agents rank each new document from their points of view. Three of these agents rank the document according to user's query, user's documents of interest in the profile, and user's group. The forth agent merges these scores and provides another score. Finally, the last agent corrects the obtained score by using previous estimated ranks and the judged ranks by the user. Using this agent that learns the users' behavior and adjusts the combination of the outputs of the other agents to estimate a rank for a document, results in a more accurate score than using each one (query, profile, or group) alone. We used LSI (Latent Semantic Indexing) method for documents and queries.