Details about the token-level labeling Task

Question

Details about the token-level labeling Task

Closed this issue 5 years ago · 3 comments

Thank you for your excellent job of the GEC task. I saw you mentioned the token-level labeling task. I didn't find it in this code. Could you give me more details about how to combine the token-level labeling task with the present work. Thank you for your amazing works again.

Answer 1 · 2019-05-09T12:17:51.000Z

Yes, the code of the token-level task hasn't merged.
It needs alignment info between source and target sentences, and we use "fast_align". For more details please refer to Figure 1 and section 4.1 of the paper.

Answer 2 · 2019-05-10T01:54:00.000Z

Thank you for your reply. Do you have any plan to release the detailed experimental codes of this part? Is the token-level task trained together with the seq2seq task ? If so, how do you choose the weight between these two tasks? Or just you trained a token-level model, and then a seq2seq model?

Answer 3 · 2019-05-15T11:56:01.000Z

Maybe later.
It's trained together.