Thanks for checking out the Gender Graph project:)

The binary model is made by word2vec:

https://github.com/dav/word2vec

Gender graph uses word embeddings trained by word2vec, the implementation is in the source repo under:

src/gender_word_plotting.c

So the source repo has a binary file in it, which seems to have been made by vec2bin:

https://github.com/sneha-belkhale/gender-word-plots

The python scripts are pretty simple. I'd like to see some methodology drill down / peer-review to make sure these models are sane. I don't really have the mathematics background myself.