0
$\begingroup$

I have a dataset of user written reviews. What I would like to do is to cluster similar reviews together.

I have already trained a word2vec model on vocabulary of 50000 words, and I have updated my data frame so that each user review has dimensions (numb_of_words, 375) where numb_of_words indicates a number of words in a given user review and 375 indicates the length of word embedding.

My question is this, can I use K-Means to cluster similar user reviews? Not individual words, but the whole reviews. If so, must I first get them all to the same dimension? Or must I perform some sort of dim. reduction technique first?

$\endgroup$
0
$\begingroup$

The usual approach with word2vec seems to be to take the average vector of each review.

Then cluster these.

Good luck - I am all but convinced that this can work particularly well, because of the underlying assumption that every point is part of a cluster in the first place... But I guess many users don't care, they just need some quantifiable result for the hype cycle.

$\endgroup$

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.