Yeah, I simply ingested words in French from a few online ballet dictionaries into a JSON object, and used a recursive context-free grammar to generate (generally) grammatically correct sentences. The result was a mixture of English and French because the dancer (Sarah Hay) requested that I add a few words of her choice in English as well.
Hey, I'm one of the creators of this -- I trained the LSTMs that wrote the computer generated segments in this film. The ballet choreography was generated with a context-free grammar, because I couldn't find enough material to make a ballet choreography LSTM.
It's worth noting that this year I used subtitle files rather than screenplays to train our LSTMs, so we only had dialogue rather than dialogue + action descriptions. The Ars Technica article explains everything.
Hey, I'm the creator, and I agree that it's a bit hit or miss. I think the definitions from the words it generates on its own every 90 minutes (these are the ones without anyone's twitter handle at the bottom) are (on average) quite a bit better than those from the words people are tweeting at it, which it typically (and somewhat expectedly) has a bit more trouble with.
This bot was really just a creative experiment, and I'm pretty new to machine learning, so I'd love to hear any specific suggestions you might have to improve it.
Hey, I'm the creator. This would probably work, you'd want to use a unique delimiter character that's not in the rest of the corpus --- so not a comma. (I'm using the pipe character.)