Sequential Latent Semantic Indexing

Sequential version of the latent semantic indexing method

Thus, this work offeres a sequential version of the LSI algorithm
(SLSI). Its main difference from the existing algorithms is that the
dimension of space is not fixed and dynamically changes to
ensure a given level of relative approximation error of a matrix of
observations. Experiments with a real text collections show that
the SLSI algorithm can be seen as a compromise, which has a
lower computational complexity and memory requirements
compared to the standard LSI method and does not lead to a
decrease of quality of classification in contrast to other sequential
algorithms.