barshana-banerjee/paraqa

A Dataset with Multiple Paraphrase Responses for Single-Turn Question Answering

paraqa's Issues

Missing Square Brackets

Hi!

Thank you for delivering such verbalized dataset !

However, I have noticed that some verbalizations come with missing square brackets around the raw answer. Overall, those samples only stand for 6.2% of the training set and 5.7% of the test set.
As an example, the training example with uid 342 has its second verbalized form with missing answer delimiters. If some raw answer could be found back from other verbalized forms, some verbalizations of same example possess different raw answers which leads to inconsistency.

A list of concerned uids:
missing_square_brackets_train.txt
missing_square_brackets_test.txt

Do you plan to correct that in future ? Do you accept any PR ?

Thank you !