Information Extractions from Transcripts

[stable mango]

Hey Deepgram team 👋
I’m looking for guidance on how to properly extract specific structured information from transcribed phone calls.

Use case:
I upload recorded phone calls (wholesale real estate conversations) and receive a transcription. From each call, I need to extract only the following fields:

How fast the prospect wants to sell

How much they’re looking for (price)

Condition of the property

Reason they want to sell

I don’t want extra or unrelated information, just these specific data points.

Questions:

Is it possible to reliably extract this kind of structured info using Deepgram?

Should this be handled via the Topics tab, Intent tab, or something else (like custom prompts or post-processing)?

If Topics or Intents are the right approach, what would those configurations look like for this use case?

Or is the recommended approach to do transcription first and then run a separate NLP/LLM step for extraction?

I want to make sure I’m using Deepgram the right way instead of forcing a workaround. Any guidance or best practices would be super helpful 🙏

Thanks in advance!

[pliant breach]

Thanks for asking your question. Please be sure to reply with as much detail as possible so the community can assist you efficiently.
-# If you haven't done so, ensure your Discord and Github profiles are linked to Deepgram so you can earn points to redeem on cool stuff just by being active!

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• The programming language you are working in (e.g. JavaScript, Python).
• The deepgram product you are using (e.g Speech to Text, Agent API)
• A request ID that triggered your error or issue.

[stable mango]

Thanks for the response!

I’m not specifying a particular language right now — I’m just using Deepgram’s default speech-to-text settings to transcribe English phone calls. After the transcription is complete, my goal is to extract specific structured information from the call (timeline to sell, asking price, property condition, and reason for selling).

Let me know if the language or model selection affects whether this is possible, or if there’s a recommended setup for this type of post-transcription extraction.

[keen sierra]

this likely wouldn't be done at the STT level. you could consider using our voice agent API with agents and tool calls to extract and store this information. or if you're just looking to do this as post-processing outside of an agent workflow, I'd recommend asking structured, step-wise questions for these pieces of information if you already already, and using an LLM to post-process (you might be able to use LLM post-processing regardless of a structured Q&A flow).