In this paper we present 'Retriever', an autonomous agent that executes userqueries and returns high quality results to the user. Retriever utilizes existing search engines to obtain the starting points for its subsequent autonomous exploration of the Web. A self-training process is conducted, in order to learn the query domain, thus increasing its efficiency. When the query domain is learned, the original query is expanded, the search strategy is reformed and the agent starts looking for the documents to be presented to its user. Relevance feedback is also utilized in order to improve performance on subsequent searches on the same query.
展开▼