1

I am trying to re-train the pre-trained google Word2Vec with bunch of domain-specified documents.

So I need a google-Word2Vec source code, however, the original page of W2V now returns error when the source-code linke clicked.

Anyone knows the other mirror repo for the google W2V?

https://code.google.com/archive/p/word2vec/ Word2Vec page http://word2vec.googlecode.com/svn/trunk/ Repo Error

snapper
  • 997
  • 1
  • 12
  • 15

2 Answers2

3

It's here now: https://github.com/dav/word2vec

Also, the implementation in Gensim is great -- details and documentation here: https://radimrehurek.com/gensim/models/word2vec.html

Gabriel
  • 587
  • 5
  • 17
  • I know the gensim is awe-some but its reform() method which is used for retraining of the embeddings doesn't support keyed_vectors, which are W2V or GloVe like major embeddings sadly :n – snapper Feb 09 '18 at 03:08
  • Ah I see what you're after. You might find the answers here to be helpful: https://stackoverflow.com/questions/35117491/is-it-possible-to-re-train-a-word2vec-model-e-g-googlenews-vectors-negative300 . But you can find a clone of the original google w2v at https://github.com/dav/word2vec if you prefer that. Hope that helps, please do accept this answer if you found it useful :) – Gabriel Feb 09 '18 at 08:41
0

You can find a full clone of that repo here on GitHub: https://github.com/svn2github/word2vec

Olesya Razuvayevskaya
  • 1,148
  • 1
  • 7
  • 10