[08] He He - Sequential Decisions and Predictions in NLP
The Thesis Review
How to Frame NLP Problems as Sequential Decision Making
In terms of your thesis, so you mentioned here framing text generation as sequential decision making. Did you want to maybe talk about other ways that you could use frame NLP problems as imitation learning and maybe going to what you did in the thesis here with like this controller and batch system? Yeah, I think whenever you have a search component, you should be able to frame at like, for example, structure prediction, then you should beable to frame that as sequential decisions. And also when you have interactions like dialogue, you would want an agent to interact with a user and depending on the user's feedback or the user's action, you might want to change what you're going to do.
Remember Everything You Learn from Podcasts
