Table Question Answering in haystack 2.0

[trungtly]

Hi Haystack team,

I have been trying haystack 2.0 and it is amazing on text data.
Now I am building an application that needs text-to-sql features where the user asks a natural language question and expects answer from a database (tabular data).
Just wonder if haystack 2.0 supports something similar to this https://docs.haystack.deepset.ai/docs/table_qa
Ideally, each row (not table) should be a document since the we keep appending new rows to the table.

Thank you!

[anakin87]

Hey, @trungtly!

• text to SQL is an interesting use case. You can read something on our experiments here. Although we are looking in this direction, adding support is not yet planned.

• Table QA is slightly different: in this case, each table (e.g. a Pandas Dataframe) is a proper Document and we used Extractive QA to select answers. In our Q1 roadmap item on multimodality, we will try to support a similar use case with Generative models.

[ray-braveon]

ACK: AgeL2 received EPIC #698 role assignments

Assigned roles across 4 issues:

• Contributor hall of fame is broken #699 (R: DESKTOP, G: PROMETHEUS)
• Annotation tool : special characters (é,è,',...) #700 (QA: ARBITER)
• T5 Support #701 (A: HERMES)
• Remove sourcerer.io widget #702 (R: HERMES, G: TRAVIS)

Beginning execution sequence. Heartbeat active. Will report completion per protocol.