search

laresbernardo / elemetrica

0 stars 1 forks source link

looking for a guinea pig #1

Open jwijffels opened 4 years ago

jwijffels commented 4 years ago

Hello, would you be interested in trying out this package: https://github.com/bnosac/doc2vec I need a few testers before putting it on cran, you seem to have used word2vec::doc2vec recently so I think you are up to this :) feel free to provide feedback

laresbernardo commented 4 years ago

Hi @jwijffels thanks for sharing! I tested the word2vec library and does exactly what I needed; but, as I'm training the models with h2o, their w2v functions run seamlessly for new predictions and pre-processing steps. Both work just fine but glad to try any new functionality! Cheers.

jwijffels commented 4 years ago

Just testing the doc2vec::paragraph2vec out on my own local data as well (using the features as input to predictive models). Would be great to have some extra testers - I see you were comparing some predictive models based on the text, maybe you could also test it alongside doc2vec instead of just averaging or doing the inner product of the word vectors. Only if you have time/interest of course. Is h2o also providing doc2vec or is it just averaging of the vectors?