1

Is there a package that supports probabilistic latent semantic analysis for R? I found the LSA package, but is there one that specifically performs pLSA? Thanks.

IRTFM
  • 258,963
  • 21
  • 364
  • 487
TomR
  • 546
  • 8
  • 19

1 Answers1

2

First hit on google search: r-project topic models probabilistic latent semantic analysis

www.jstatsoft.org/v40/i13/paper

Edit: The question probably should have been closed, and my answer was probably more appropriate as a comment, but it did get two comments from @rarry with greater knowledge than I possess. Since I think answer content gets more "search attention" than comments I'm going to incorporate the comments here;

As Bob Quek already mentioned, the answer is incorrect. The paper is about quite different model, namely LDA. LDA differes from PLSA in that, in LDA we introduce prior over parameters that describe distribution of words in topics and distribution of topics in document. In PLSA we have no priors and parameters are estimated directly. In other words, LDA i a bayesian version of PLSA. – Jul 9, 2017 at 10:01

You can do PLSA analysis using svs package https://cran.r-project.org/web/packages/svs/index.html – Jul 9, 2017 at 10:14

If rarry wants to compose an answer, I'd be happy to delete mine (although accepted answers might not be delete-able, since I know that Q's with accepted answers are not close-able.).

IRTFM
  • 258,963
  • 21
  • 364
  • 487
  • Thanks. I have no idea why I couldn't find that...perhaps my firm's internet constraints. In any case, thanks. – TomR Aug 20 '12 at 14:17
  • 2
    The paper says nothing about PLSA though. It only talks about latent dirichlet allocation and correlated topics models. – Bob Quek Aug 09 '13 at 21:05
  • I should have voted to close this last year. Requests for finding tools to do X are off-topic on SO. – IRTFM Aug 09 '13 at 22:04
  • 1
    As Bob Quek already mentioned, the answer is incorrect. The paper is about quite different model, namely LDA. LDA differes from PLSA in that, in LDA we introduce prior over parameters that describe distribution of words in topics and distribution of topics in document. In PLSA we have no priors and parameters are estimated directly. In other words, LDA i a bayesian version of PLSA. – rarry Jul 09 '17 at 10:01
  • 2
    You can do PLSA analysis using svs package https://cran.r-project.org/web/packages/svs/index.html – rarry Jul 09 '17 at 10:14