Soniox Docs » Omnio » Audio Capabilities » Information Extraction

Information Extraction

Omnio can be used to extract various information from the audio with ability to follow detailed instructions.

Prompt

Extract all numerical information from the audio.
Output the extracted data as JSON in this format:
[{"number": ..., "description": ...}, ...]

[
  {
    "number": 12345,
    "description": "order number"
  },
  {
    "number": 24,
    "description": "hours to deliver the package"
  },
  {
    "number": 20.45,
    "description": "price of the package in USD"
  }
]

Prompt

Extract all people mentioned and their roles.
For each topic discussed, list one quote and include its timestamp range.
Output the information in JSON format like this:
{
  "people": [
    {
      "name": ...,
      "role": ...
    }
  ],
  "topics": [
    "<topic_name>": {
      "quote": ...,
      "timestamp_range": "MI:SS-MI:SS"
    }
  ]
}

{
  "people": [
    {
      "name": "Mr. Smith",
      "role": "Customer"
    },
    {
      "name": "Mr. Jones",
      "role": "Customer Support Representative"
    }
  ],
  "topics": {
    "Product Issue": {
      "quote": "I got a laptop and every time I plug it in, it's not working.",
      "timestamp_range": "00:07-00:10"
    },
    "Order Verification": {
      "quote": "Could you tell me your order number please?",
      "timestamp_range": "00:45-00:47"
    },
    "Discount Explanation": {
      "quote": "We can give you an additional 10% discount.",
      "timestamp_range": "01:40-01:43"
    }
  }
}