Document clustering using an inverted file approach

An automatic document clustering procedure is described which does not require the use of an inter-document similar ity matrix and which is independent of the order in which the documents are processed. The procedure makes use of an initial set of clusters which is derived from certain of the terms in the indexing vocabulary used to characterise the documents in the file. The retrieval effectiveness obtained using the clustered file is compared with that obtained from serial searching and from use of the single-linkage clustering method.