Interactive document retrieval with relational learning and it's application to web search engine

摘要

This paper describes an interactive information retrieval system which mediates between the user and the web search engine to realize relevance feedback. This system automatically generates a query for search engine by converting rules which are learned to distinguish between relevant pages and unrelevant ones user has already seen and judged during retrieval. The rules consist of keywords, operators, and region restriction to represent the feature of the web page. They are learned by the algorithm which adopts separate-and-conquer strategy and top down heuristic search using information gain.