Named entity recognition and automatic categorization

Hi Lennart,

There is no simple way at the moment for the interpolator to do this in one request for two fields. Currently it only supports many-context-to-one-field or one-context-to-one field.

What you can do is either:

• that you create two different taxonomy fields for each of the vocabularies and then add a rule for each. This of course unfortunately means almost twice the cost for OpenAI queries.
• or put both vocabularies in the same field

Hope that answers the question.

Thank you!

Is this something OpenAI's Function calling could help achieve down the line?

And will AI Interpolator eventually be able to "parse" structured data returned as json?

So it already work completely by parsing structured JSON data, basically anything that gets generated is returned as JSON. That's how it can work with complex fields like custom fields (see: https://workflows-of-ai.com/workflow/custom-complex-fields).

This is superior to function calling for a no-code GUI solution, since the function calling needs you to do a very good spec of inputs and outputs. In this case we use one-shot learning to generate the (JSON) output, which is easy enough for an end user to generate and good enough to handle complex data.

This won't however solve the problem of wanting to create multiple fields with the same prompt. That one we have a solution for in our company, but we are not ready to open source it yet.

Similar request ✨ Could you implement a JSON parser feature? Active

Thank you very much for the detailed answer, Marcus.