Reports the results of experiments in document clustering using three well-known test collections. (Author)

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 logic of different types of weighting are discussed, and experiments testing weighting schemes of these types are described. The results show that one type of weighting leads to material performance improvements in quite different collection environments. (22 references) (Author)

Heuristic methods for the construction of term classes are presented and experimental results are obtained to illustrate the usefulness of the method. (Author/PF)

An automatic classification technique using patterns formed by citations in document bibliographies was found to give 62 percent precision and 45 percent recall in a sample file of physics documents. (PF)

Patents are clustered by similar cited-by" patterns using MacQueen's algorithm. The clusters are enriched by the addition of citing patents. These maximally connected clusters, or cliques, form the classification. (27 references) (Author)

Reports on a series of programs that have been developed to process data bases, consisting of textual items, and to index and arrange (classify) the data items in accordance with an automatically generated classification system. (Author)