Squalar Spectral Map is a simple package that takes a term-document matrix, calculates a term co-occurrence matrix, calculates its eigendecomposition with SVD, and finally maps terms to the visible spectrum by comparing a term's co-occurrence vector to the eigenvectors.
Java
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
doc
figures
src
.classpath
AUTHORS
LICENSE
README

README

Squalar Spectral Map is a simple package that takes a term-document matrix, calculates a term co-occurrence matrix, calculates its eigendecomposition with SVD, and finally maps terms to the visible spectrum by comparing a term's co-occurrence vector to the eigenvectors.

Third-party work includes Svdlib by Adrian Kuhn and David Erni, and an image of the visible spectrum by Deborah S Krolls (http://commons.wikimedia.org/wiki/File:Spectrum4websiteEval.png).

Two files are needed to create the document spectrum:
1) Libsvm-formatted sparse term-document matrix.
2) Ordered list of index terms.

The class org.squalar.spectralmap.Decompose will perform the decomposition (see the javadoc for parameters).

Based on the decomposition, the class org.squalar.spectralmap.TermSpectrum will create a png image of the spectrum of a given term (see the javadoc for parameters).

For more details and publications, visit http://squalar.org.