Fix Datasets Currently Mapped as Tasks.TEXTUAL_ENTAILMENT

[jason-fries]

Several datasets are currently tagged as supporting Tasks.TEXTUAL_ENTAILMENT incorrectly. Perhaps I'm misunderstanding the tasks, but these largely seem like text classification/labeling problems not entailment.

• medical_data is a sentiment analysis task not textual entailment. Better suited to pairs or text classification
• evidence_inference this is labeling a relationship between text snippets i.e., "intervention of interest either significantly increased, significantly decreased or had significant effect on the outcome, relative to the comparator"

Fix is migrating these to the correct schema

[shamikbose]

@jason-fries The issue with medical_data is addressed in the conversation in #613

So every text comes with a drug mention and what the text thinks of that specific drug. Putting this in the Classification format loses some of that information

[jason-fries]

@shamikbose thanks for the comment. If a simple classification schema isn’t suited then this should be a text pairs task if it involves reasoning over 2 units of text. The most important issue is that this is not an entailment task.

[shamikbose]

I can tackle the medical_data tomorrow

[shamikbose]

@jason-fries medical_data is updated to bigbio_pairs in #684

[phlobo]

Both datasets mentioned here now have other tasks assigned, therefore I will close the issue.