Describes a computerized information retrieval system that uses three agglomerative hierarchic clustering algorithms--single link, complete link, and group average link--and explains their implementations. It is noted that these implementations have been used to cluster a collection of 12,000 documents. (LRW)

Describes a cluster-based information retrieval procedure that can significantly reduce the computational requirements of the single linkage method, while still maintaining the retrieval effectiveness of the resulting classifications. Use of nearest neighbors, experimental details, and results and conclusions are highlighted. Fourteen references…

The objectives of this paper are to describe the effect of using weighted index terms in a document retrieval system, and to evaluate retrieval performance when queries are expanded by terms occurring in clusters with the query terms. (16 references) (Authors)

Substantial alterations to a system often have little or no effect on particular collections. This may be due to poor separation of relevant and non-relevant documents. The paper presents a procedure for characterizing this separation, which can show whether proposed modifications of the base system are likely to be useful. (8 references)…

Retrieval performance with automatic term classifications for three test collections has been variable. This paper attempts to discover why. The real difference between the collections is in the separation of relevant from non-relevant documents. The separation is so poor that classification cannot be expected to succeed. (14 references)…

The objectives of this paper are to describe the effect of using weighted index terms in a document retrieval system, and to evaluate retrieval performance when queries are expanded by terms occurring in clusters with the query terms. Three data collections, each indexed by several methods, two of which were studied and reported on in previous…

Examines research on the observed values of retrieval effectiveness obtained in searches of files of nearest neighbor document clusters. Results show interdocument similarities used to generate nearest-neighbor clusters are significantly different from randomly generated clusters. (Contains 14 references.) (EAM)

An evaluation of graph theoretical clusters of index terms which can be extracted from an automatically indexed document collection, and the effects of employing such clusters in automatic document retrieval are described. (19 references) (Author)

Introduces a novel approach to unsupervised text summarization. Proposes an "information-centric" approach to evaluation, where the quality of summaries is judged not in terms of how well they match human-created summaries but in terms of how well they represent their source documents in information retrieval tasks such as document…

Reports on comparative study of document classifications produced by use of single linkage, complete linkage, group average, and Ward clustering methods. Findings of work that compares use of clusters consisting of pairs of documents with conventional best match searches are also reported. Thirty-four references are provided. (EJS)

Considers classifications produced by application of single linkage, complete linkage, group average, and word clustering methods to Keen and Cranfield document test collections, and studies structure of hierarchies produced, extent to which methods distort input similarity matrices during classification generation, and retrieval effectiveness…

The effectiveness of using the age of references to control the exhaustivity of the reference representation in information retrieval was investigated through analysis of optimal cluster-based retrieval results. The results show that the foreground representation at its optimal level of exhaustivity is restricted to references with ages less than…

Using examples of data from the areas of information retrieval and of multivariate data analysis, six hierarchic clustering algorithms (single link, median, centroid, group average, complete link, Wards's) are examined and evaluated by using three proposed coefficients of hierarchic structure. Nine references are cited. (EJS)

A measure for the quantification of the changes in classification under small changes in data is proposed. (Author)

The author emphasized that one cannot conclude from the experiments reported upon that term clusters (or equivalently, keyword classifications or thesauruses) are not useful in retrieval. (2 references) (Author)