Spectral co-clustering is a generic method of computing co- clusters of relational data, such as sets of documents and their terms. Latent semantic analysis is a method of document and term smoothing that can assist in the information retrieval process. In this article we ex- amine the process behind spectral clustering for documents and terms, and compare it to Latent Semantic Analysis. We show that both spectral co-clustering and LSA follow the same process, using different normal- isation schemes and metrics. By combining the properties of the two co-clustering methods, we obtain an improved co-clustering method for document-term relational data that provides an increase in the cluster quality of 33.0%.

A matrix D, of pairwise dissimilarities between m row objects and n column objects, can be clustered: amongst row objects or column objects; amongst the union of row and column objects; and amongst the union of row and column objects containing at least one object of each type (co-clusters). The coVAT algorithm, which builds images for visual assessment of clustering tendency for these problems, is limited to mn O(10^4*10^4). We develop a scalable version of coVAT that approximates coVAT images when D is very large. Two examples are given to illustrate and evaluate the new method.

Latent semantic analysis (LSA) is a generalised vector space method (GVSM) that uses dimension reduction to generate term correlations for use during the information retrieval process. We hypothesised that even though the dimension reduction establishes correlations between terms, the reduction is causing a degradation in the correlation of a term to itself (self-correlation). In this article, we have proven that there is a direct relationship to the size of the LSA dimension reduction and the LSA self-correlation. We have also shown that by altering the LSA term self-correlations we gain a significant increase in precision during the information retrieval process.

Probabilistic latent semantic analysis (PLSA) is a method for computing term and document relationships from a document set. The probabilistic latent semantic index (PLSI) has been used to store PLSA information, but unfortunately the PLSI uses excessive storage space relative to a simple term frequency index, which causes lengthy query times. To overcome the storage and speed problems of PLSI, we introduce the probabilistic latent semantic thesaurus (PLST); an efficient and effective method of storing the PLSA information. We show that through methods such as document thresholding and term pruning, we are able to maintain the high precision results found using PLSA while using a very small percent (0.15%) of the storage space of PLSI.