We present a detailed analysis of matrices satisfying the so-called low-ran
k-plus-shift property in connection with the computation of their partial s
ingular value decomposition (SVD). The application we have in mind is laten
t semantic indexing for information retrieval, where the term-document matr
ices generated from a text corpus approximately satisfy this property. The
analysis is motivated by developing more efficient methods for computing an
d updating partial SVD of large term-document matrices and gaining deeper u
nderstanding of the behavior of the methods in the presence of noise.