Add Text-to-SQL progress (Dialogue)

[willyspinner]

I've been reading up on Text-to-SQL problems (given natural language question & DB schema, generate a legal SQL query) and they seem to be an interesting subfield of Dialogue Answering Systems.

There are multiple datasets/ challenges that track their SOTAs, for example:

1. The spider challenge
2. The CoSQL challenge
3. The SparC challenge

all of which have interesting tasks besides the standard Text-to-SQL problem, like Natural Language Response Generation given SQL and Query execution Results.

I realize that the spider dataset is under the semantic-parsing part, but SParC and CoSQL are more challenging datasets that incorporate multi-turns, context, as well as ambiguous and unanswerable questions that may present to be harder than the spider challenge.

I was wondering if we could add them?

[sebastianruder]

That's a great idea! I think it may be potentially clearer to add them to the Semantic Parsing area as they are conceptually more similar to the Spider dataset than the standard dialogue setting. What do you think?