Information retrieval based on fuzzy reasoning

摘要

This paper discusses modelling, design and implementation issues for a knowledge-based information retrieval system. The system manages natural languages queries and returns a ranked set of relevant documents computing for each of them a degree of support depending on multiple sources of evidence. In our approach, the retrieval process is regarded as the problem of determining an implication relationship between a document and a query and assessing the plausibility of the implication.In the following, after a brief remark about the need for modeling the inherent imprecision associated with the retrieval process, a model based on fuzzy reasoning is proposed. Then the architecture is presented of a prototype system which has been implemented to evaluate the improvement in the overall retrieval effectiveness. Finally, we discuss implementation details with reference to search strategies that have been developed to reduce the impact of using imprecise knowledge upon the computational efficiency of the system.