Natural language processing is in a curious place right now. It’s not immediately obvious how successful it’s been, or how close the field is to viable, production-ready techniques (in the same way that, say, computer vision is).
Let’s talk about Bayesianism. It’s developed a reputation (not entirely justified, but not entirely unjustified either) for being too mathematically sophisticated or too computationally intensive to work at scale.
Recently I’ve started using PyMC3 for Bayesian modelling, and it’s an amazing piece of software! The API only exposes as much of heavy machinery of MCMC as you need - by which I mean, just the pm.sample() method.
I was recently inspired by this following PyData London talk by Vincent Warmerdam. It’s a great talk: he has a lot of great tricks to make simple, small-brain models really work wonders, and he emphasizes thinking about your pr...