TopicScore: The Topic SCORE Algorithm to Fit Topic Models

Provides implementation of the "Topic SCORE" algorithm that is proposed by Tracy Ke and Minzhe Wang. The singular value decomposition step is optimized through the usage of svds() function in 'RSpectra' package, on a 'dgRMatrix' sparse matrix. Also provides a column-wise error measure in the word-topic matrix A, and an algorithm for recovering the topic-document matrix W given A and D based on quadratic programming. The details about the techniques are explained in the paper "A new SVD approach to optimal topic estimation" by Tracy Ke and Minzhe Wang (2017) <arXiv:1704.07016>.

Version: 0.0.1
Depends: R (≥ 3.5.0)
Imports: utils, stats, graphics, RSpectra, combinat, quadprog, methods, Matrix, slam
Published: 2019-06-06
Author: Minzhe Wang [aut, cre], Tracy Ke [aut]
Maintainer: Minzhe Wang <minzhew at>
License: MIT + file LICENSE
NeedsCompilation: no
CRAN checks: TopicScore results


Reference manual: TopicScore.pdf


Package source: TopicScore_0.0.1.tar.gz
Windows binaries: r-devel:, r-release:, r-oldrel:
macOS binaries: r-release (arm64): TopicScore_0.0.1.tgz, r-release (x86_64): TopicScore_0.0.1.tgz, r-oldrel: TopicScore_0.0.1.tgz


Please use the canonical form to link to this page.