search

ShannonAI / mrc-for-flat-nested-ner

Code for ACL 2020 paper `A Unified MRC Framework for Named Entity Recognition`
643 stars 117 forks source link

How did you handle "no answer" scenarios? #86

Open seanswyi opened 3 years ago

seanswyi commented 3 years ago

I'm aware that usually in question answering if the model deems that there's no answer for a question (or a question is "impossible") then the position of the [CLS] token is output. However, I've noticed that that doesn't seem to be the case for your work. How did you go about taking care of "no answer" scenarios? Thanks.

yangjing9264 commented 2 years ago

I'm aware that usually in question answering if the model deems that there's no answer for a question (or a question is "impossible") then the position of the [CLS] token is output. However, I've noticed that that doesn't seem to be the case for your work. How did you go about taking care of "no answer" scenarios? Thanks.

我也好奇这点，请问你现在清楚了吗