Outline of a dynamic self-tuning and adaptive information retrieval system

Abstract

A self-tuning adaptive information retrieval system as an extension of the concept of a "classical" document retrieval system, is outlined. This system accepts documents and search requests in natural language, as well as the system-proposals previously produced by the system itself or prepared by the system operator. It produces a system-proposal that consists of a list of documents ranked according to their relevance to the query.Incorporated into the system is a system valuation subsystem that uses weighted relevance judgements. This subsystem gives as output an effectiveness value and an efficiency value: both together measure the quality of an information retrieval system.The computation of the quality values and the values themselves are independent of a specific implementation. The retrieval process in this system consists of two parts, namely a query-document match and a query-query match.