multi-sense-skip-gram-clone

MSSG is a very useful form of Word2Vec that learns multiple embeddings for each word. In general MSSG has achieved superior to standard Word2Vec. Ufortunately, there are very few production ready implmentations of MSSG. This is an attempt to get the code of the Multi Sense Skip Gram model from the article "Efficient Non-parametric Estimation of Multiple Embeddings per Word in Vector Space" up to date and (hopefully) refactored so that it's easy to integrate into Scala/Java data pipelines. Original code taken from jeevan_shankar's bitbucket. Any help is appreciated so feel free to fork and make pull requests.