1
$\begingroup$

I am confused on one issue of topic modeling analysis. When we perform the topic modeling analysis over a collection of documents, do we need to provide a set of topic candidate list as the input for the analysis? Or the resulting topics are just extracted from the input documents themselves?

$\endgroup$
1
$\begingroup$

The resulting topics are just extracted from the input documents themselves.

Document are distributed over topics, and topics are distributed over words, so you can explain topics by top-words.

$\endgroup$
  • $\begingroup$ Thanks for the response. If I have a pre-existing topic list, and would like to ensure that all of the generated topic come from this list? Can LDA incorporate this kind of constraint or I have to use other algorithms? $\endgroup$ – user3269 Apr 19 '12 at 14:32
  • $\begingroup$ If you pre-existing topic are just words, you can refer Tag LDA or Labeled LDA, just google it. As far as I known, in LDA you can't do that, even you pre-set the topics(distribution over words), it will converges with new topics set. But if you have topics already, why do you need LDA? or you can just fit document to topics? $\endgroup$ – changsheng Apr 20 '12 at 1:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.