Is that allowed to use other training data?

Question

Is that allowed to use other training data?

LegendaryDan opened this issue 5 years ago · 2 comments

(1) Can we use other QA or MRC training data?
(2) Can we use the training data of non-QA tasks?

Thanks.

Answer 1 · 2019-04-30T07:06:22.000Z

Hello and thank you for the interest! This task is not officially released yet. Please wait for more news later this week.

Answer 2 · 2019-05-02T19:10:21.000Z

Hi @LegendaryDan ,

More details are now available at https://mrqa.github.io/shared.

To answer your questions:
(1) Just the official training sets.
(2) Unlabelled data is allowed. As are off-the-shelf linguistic annotation tools. For other annotated datasets that are not QA datasets, please contact us first. (Post an issue.) We hope to release a whitelist of auxiliary datasets one might consider using.