I actually began my journey in machine learning with a cloud scale natural language processing system many years ago (Java MLPs!) and at various times have worked with classification, sentiment analysis, and text generation. I’ve been looking for opportunities to play around with the current SOTA in language models and saw this fascinating paper the other day: Learning to Speak and Act in a Fantasy Text Adventure Game. I don’t often have the time to read all the way through a paper, but this one was fascinating to me for its methods as well as relevance to my principle research.
As will become apparently in future posts, in addition to being a software dork, I am also an RPG dork, so bringing the two sides together is always fun. Phil and I are on a pretty tight timetable for CHIPLAY so its unlikely we’ll get to implement anything from this paper for at least a few months, but I think in the long run it may be really valuable in our lexical analysis when trying to extract content for maps of belief space. For more details on our current work stay tuned or check out http://www.viztales.com or https://phifel.com/
ML Dork 