Using Word2vec output as Input for LDA in topic modelling

Question

I have read that the most common technique for topic modeling (extracting possible topics from the text) is Latent Dirichlet allocation (LDA). but recently I learned about another model lda2vec. However, I am interested in whether it is a good idea to try Word2Vec output as an input to LDA.

Do you think it makes sense to follow this approach for the sake of some research? As I am doing topic modeling and need some novel approach.

Answer 1

That is a good idea, and there are already some papers. I suggest you search "word embedding + LDA" instead of word2vec. Interestingly, even David Blei himself (the inventor of LDA) said he was considering doing something like this in a recent interview.