Word2Vec algorithm

User 1241 | 1/27/2015, 10:01:17 PM

hi, I am fairly new with graphlab and currently experimenting it. I heard graphlab has deeplearning feature and I wonder does it support Word2Vec type of algorithm? If you can point me to the right direction, that would be appreciated. Thank you.

Comments

User 1190 | 1/27/2015, 11:20:16 PM

Hi,

I just want a clarification of your question. Do you want to implement Word2Vec or do you want to use some existing word2vec function?

Thanks, Jay


User 1241 | 1/28/2015, 3:50:29 PM

Hi Jay, Thanks for the quick reply. I am wondering that the deep learning feature of graphlab supports word2vec. If not, what would be the easy way to implement it? using graphlab or other MLLibs such as the google code?

https://code.google.com/p/word2vec/

Thanks. Kevin


User 1190 | 1/28/2015, 6:41:09 PM

Hi Kevin,

The current deeplearning feature in Graphlab Create supports building classifiers using convolutional network and fully connected network. It does not quite have the support for building a neuralnet language model.

If you want to implement using Google's word2vec, please take a look at the SDK page: https://github.com/graphlab-code/GraphLab-Create-SDK

It will help you to using the existing word2vec implementation and make it work with GraphLab Create in python. This could be a very interesting project, and please let me know if you are excited about doing it. I will be happy to offer my support.

Jay


User 1241 | 1/28/2015, 7:03:35 PM

Hi Jay,

Thanks for the info. I will evaluate the level of effort and definitely reach out if needed.

Kevin