How to annotate our own custom dataset ?

[bvy007]

Hi,

Could you please suggest me some ideas of creating a Questioning and Answering Dataset just like SQuAD but with common set of questions for every paragraph and a specific answer from the paragraph?? Any leads would be appreciated.

[ayushjain1144]

You can checkout https://github.com/samdash/QuestionAnswers

It provides interface to enter question answers and converts into BERT format. There are some bugs in this, but served the purpose in my case. It does not athutomatically calculates the start index though. I wrote a small script to calculate and feed it. Also it won't be very difficult to just modify it to automatically take the start index