The Best-Match Problem in Document Retrieval.

This article analyzes the best-match problem in document retrieval. One major problem encountered in clustering for the purpose of reducing search time is the selection of an appropriate cluster representative. These representatives are to somehow characterize the cluster and discriminate it from other clusters, so that a search may be able to find the nearest cluster by comparing each cluster representative with the incoming query. Two different approaches to the selection of a typical member are possible. One is through the definition of typicality. A cluster representative should be such that an incoming query will be diagnosed into the cluster containing the documents relevant to the query. Unfortunately, as it stands this is of very little help in choosing the cluster representative since the relevant documents are normally not known in advance. However, it is reasonable to suppose that a query identical to a document, that is, a document treated as a query, should retrieve that document. Experiments with this kind of cluster representative have been reported elsewhere.