119 - Social NLP, with Diyi Yang
NLP Highlights
00:00
How to Evaluate a Model for Identifying Subject to Bias
We use fluency, mini preserving and style related matters to evaluate the model. The most important matter here is whether giving sentence is less bias or more bias. In other case, annotators may not be qualified to work on our annotation. So only people who data very well in the trinity can work on the actual evaluation part. I think what is meaning that the self is a very interesting question. We have an ongoing collaboration with Virginia University on NLP as we probably shouldn't throw around these like words without thinking about what does that actually mean?
Transcript
Play full episode
