Fuzzy tolerance relations and relational maps applied to information retrieval

One of the major problems in automatic indexing and retrieval of documents is that usually it cannot be guaranteed that the user queries include (all) of the actual words that occur in the documents that should be retrieved. Also it often happens that words with several meanings occur in a document, but in a rather different context from that expected by the querying person. In order to achieve better recall and higher precision, fuzzy tolerance and similarity relations have been introduced based on the counted or estimated values of (hierarchical) co-occurrence frequencies. This study addresses the problem of how these relations can be generated from the occurrence frequencies, especially as these are based on possibilistic rather than probabilistic measures, and also how the relations can be implemented by fuzzy relevance matrices. (C) 2002 Elsevier Science B.V. All rights reserved.