139 - Coherent Long Story Generation, with Kevin Yang
NLP Highlights
00:00
How to Improve Evaluation for Long Text in General
In our follow-up paper we compared passages of stories against other passages of stories which corresponded to same parts of the outline. We used a different evaluation platform this like it's like surge AI platform instead of mechanical Turk where on average the you like annotators are just like higher quality so in one solution is just like get better annotators what we did another thing you can do is to try to like if possible you compose the stories into shorter chunks. If you can like pair them up in this way you can get a you can like if the annotator doesn't have to read quite as much and can just focus on these like shorter passages It makes it a little easier as it has been
Transcript
Play full episode
