Survey of Text Mining Clustering, Classification, and Retrieval Scanned by Velocity

The vector space information retrieval system, originated by Gerard Salton [Sal71, SM83J, represents documents as vectors in a vector space. The document set comprises an m x n term-document matrix A, in which each column represents a document, and each entry A(i, j) represents the weighted frequency of term i in document j. A major benefit of this representation is that the algebraic structure of the vector space can be exploited [BDO95J. To achieve higher efficiency in manipulating the data, it is often necessary to reduce the dimension dramatically. Especially when the data set is huge, we can assume that the data have a cluster structure, and it is often necessary to cluster the data [DHS01] first to utilize the tremendous amount of information in an efficient way. Once the columns of A are grouped into clusters, rather than treating each column equally regardless of its membership in a specific cluster, as is done in the singular value decomposition (SVD) [GV96], the dimension reduction methods we discuss attempt to preserve this information.