I tried it out on one of the Udacity Deep learning assignments using the Wasserstein loss functions built into tensorflow. I was unsuccessful in my limited use. The discriminator always ‘won out’ rather than the combo finding a saddle point. I eventually got my project to work without it, and did not go back to compare against just swapping EM back in.
The one that everyone uses from Python isn't the easiest thing to install, doesn't have a great API and isn't easy to extend.
I think Gensim recently added it, but I think they use the same backend solver.
Edit: this is a better article on EMD anyway: https://markroxor.github.io/gensim/static/notebooks/WMD_tuto...
Edit 2: I forget Textacy has an implmentation built on Spacy. Still uses the same backend solver, but the API is nice (https://chartbeat-labs.github.io/textacy/api_reference.html#...)
Currently it's unfeasible to implement even simple real time search engine using it.