Speeding Up Latent Semantic Analysis: A Streamed Distributed Algorithm for SVD Updates

Warning

This publication doesn't include Faculty of Arts. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

ŘEHŮŘEK Radim

Year of publication 2010
Type Article in Proceedings
Conference Proceedings of the 3rd International Conference on Agents and Artificial Intelligence (ICAART)
MU Faculty or unit

Faculty of Informatics

Citation
Web http://www.icaart.org/Program_Saturday.htm
Field Information theory
Keywords svd lda lsi
Description The purpose of Latent Semantic Analysis (LSA) is to find hidden (latent) structure in a collection of texts represented in the Vector Space Model. LSA was introduced in~\cite{deerwester1990indexing} and has since become a standard tool in the field of Natural Language Processing and Information Retrieval. At the heart of LSA lies the \emph{Singular Value Decomposition} algorithm, which makes LSA (sometimes also called Latent Semantic Indexing, or LSI) really just another member of the broad family of applications that make use of SVD's robust and mathematically well-founded approximation capabilities, from Image Processing; or Signal Processing, where SVD is commonly used to separate signal from noise. SVD is also used in solving shift-invariant differential equations, in Geophysics, in Antenna Array Processing, \ldots}. In this way, although we will discuss our results in the perspective and terminology of LSA and Natural Language Processing, our results are in fact applicable to a wide range of problems and domains across much of the field of Computer Science.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.