105 - Question Generation, with Sudha Rao
NLP Highlights
00:00
Is There a Better Way to Model This Utility?
The question that you actually see in the data is also dependent on what the answerer knows, right? There is a lot of tisuse specific knowledge that cannot be modelled or is completely missing here. The idea was using sik to seek models at a lower level to generate the question given a context. And we also had an answer generator, which would generate an answer given a context and a question. Both these were sick to seek models trained using emily objective function. Now, once you the answer, the next part would be to understand whether this answer increases the utility of the original post or not. So for that, what we did is used reinforcement learning approach where we considered
Transcript
Play full episode
