next up previous
Next: Singular Value Decomposition Up: Sparse Matrix Reordering Title Previous: Introduction

2. Background

  The initial phase of most vector space information retrieval models such as Latent Semantic Indexing [DDF+90,FD92], involves the construction of a term-by-document matrix. Each element of a term-by-document matrix reflects the occurrence of a particular word in a particular document, i.e.,

 

where is the number of times or frequency in which term i appears in document j. As one does not expect that each word will appear in every document, the matrix A is typically sparse with rarely any noticeable nonzero structure. As discussed in [Dum91], local and global weightings can be applied to either increase or decrease the importance of terms within or among documents so that each element may be cast as

 

where is the local weighting for term i in document j, and is the global weighting for term i.





Michael W. Berry (berry@cs.utk.edu)
Mon Jan 29 14:30:24 EST 1996