Currently, most approaches to retrieving textual materials from scient
ific databases depend on a lexical match between words in users' reque
sts and those in or assigned to documents in a database. Because of th
e tremendous diversity in the words people use to describe the same do
cument, lexical methods are necessarily incomplete and imprecise. Usin
g the singular value decomposition (SVD), one can take advantage of th
e implicit higher-order structure in the association of terms with doc
uments by determining the SVD of large sparse term by document matrice
s. Terms and documents represented by 200-300 of the largest singular
vectors are then matched against user queries. We call this retrieval
method latent semantic indexing (LST) because the subspace represents
important associative relationships between terms and documents that a
re not evident in individual documents. LSI is a completely automatic
yet intelligent indexing method, widely applicable, and a promising wa
y to improve users access to many kinds of textual materials, or to do
cuments and services for which textual descriptions are available. A s
urvey of the computational requirements for managing LSI-encoded datab
ases as well as current and future applications of LSI is presented.