Use Noise Contrastive Esimation (NCE) for word2vec

[weka511]

Mikolov et al, Distributed Representations of Words and Phrases and their Compositionality, recommend creating negative samples, k for each central word, where k is 5-20 for small datasets, 2-5 for large.

[weka511]

Fixed for word2vec2.py, still need to do word2vec.py

[weka511]

Sort out terminology, NCE vs NEG

[weka511]

Working