Dealing with no answer?

[search4mahesh]

Hi,

Dataset is more like SQUAD 1.0 where it always tries to find answer. SQUAD 2.0 solved that problem. Is there any way we can solve this for DROP dataset?

Thanks Mahesh

[wjdghks950]

@search4mahesh I think you need to implement a different head for classifying the answerability of a given question (e.g., answerable or not). Current models on DROP (including this one) do not have such capability. Moreover, the questions in DROP have the ability to count, sort and generate answers based on the numbers appearing in the passage.