Open speakerjohnash opened 7 years ago
Currently we're using Gensim for word2vec. However Gensim doesn't give us access to the internals of the algorithm. In order for our word embeddings to better match the meaning of users on Prophet we should build our own word2vec training module.
Currently we're using Gensim for word2vec. However Gensim doesn't give us access to the internals of the algorithm. In order for our word embeddings to better match the meaning of users on Prophet we should build our own word2vec training module.