ASR transcript schema

Format JSON transcripts correctly to ensure accurate processing.

When Medallia Speech ingests audio via the Speech API, we create a transcript in JSON format containing utterances and attributes that are used by Text Analytics and Reporting. All metrics and insights are calculated from that transcript rather than directly from the audio.

If you wish to provide your own complete JSON transcript rather than raw audio or video, you must be careful to format the transcript exactly as the Speech pipeline would have formatted it, including all necessary attributes, so that services further along the pipeline can work with the data.

Example transcript

For your convenience, you can copy the template below or download an example JSON file.

{
  "request_id": "074b5760-a128-420f-b7e6-1cevn098765c",
  "confidence": 0.91,
  "duration": 166.38,
  "channels": 2,
  "utterances": [
    {
      "confidence": 0.88,
      "end": 7.03,
      "start": 5.85,
      "channel": 0,
      "speakerId": "spk_0",
      "emotion": null,
      "sentiment": null,
      "text": "How can I help you?",
      "events": [
        {
          "confidence": 0.95,
          "end": 6.01,
          "start": 5.85,
          "word": "How"
        },
        {
          "confidence": 0.93,
          "end": 6.19,
          "start": 6.01,
          "word": "can"
        },
        {
          "confidence": 0.92,
          "end": 6.27,
          "start": 6.19,
          "word": "I"
        },
        {
          "confidence": 0.94,
          "end": 6.91,
          "start": 6.71,
          "word": "help"
        },
        {
          "confidence": 0.91,
          "end": 7.03,
          "start": 6.91,
          "word": "you?"
        }
      ]
    },
    {
      "confidence": 0.92,
      "end": 70.56,
      "start": 69.96,
      "channel": 1,
      "speakerId": "spk_1",
      "emotion": null,
      "sentiment": null,
      "text": "Good morning.",
      "events": [
        {
          "confidence": 0.90,
          "end": 70.14,
          "start": 69.96,
          "word": "Good"
        },
        {
          "confidence": 0.92,
          "end": 70.30,
          "start": 70.14,
          "word": "morning."
        }
      ]
    }
  ],
  "language": "en_us"
}

Schema attributes

The following table describes the attributes that must be present in the schema:


Attribute	Data type	Description	Notes
request_id	string	Unique identifier for the transcription request
confidence	number (double)	Overall model confidence score for the transcription, typically between 0 and 1
duration	number (double)	Total duration of the audio, in seconds	Required to populate the duration field
channels	integer	Number of audio channels present in the source file
utterances	array<utterance>	Ordered list of utterances (speech segments) that make up the transcription
utterances.confidence	number (double)	Model confidence score for this utterance	Required to calculate agent and customer clarity fields
utterances.end	number (double)	End time of the utterance, in seconds, from the beginning of the audio	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.start	number (double)	Start time of the utterance, in seconds, from the beginning of the audio	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.channel	integer	Audio channel on which this utterance was detected
utterances.gender	string	The gender of the speaker	Required to populate agent and customer gender fields Not available with Engine 2 languages
utterances.speakerId	string	Identifier of the speaker attributed to this utterance	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.emotion	string	Emotion detected for this specific utterance Accepted values: positive mostly_positive neutral mostly_negative negative	Required to populate agent and customer emotion fields `emotion` must be `true` in the API request Not available with Engine 2 languages
utterances.sentiment	string	Sentiment detected for this specific utterance Accepted values: positive mostly_positive neutral mostly_negative negative mixed
utterances.text	string	The transcribed text of the utterance
utterances.events	array<event>	Ordered list of word-level events within this utterance
utterances.events.confidence	number (double)	Model confidence score for this word	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.events.end	number (double)	End time of the word, in seconds, from the beginning of the audio	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.events.start	number (double)	Start time of the word, in seconds, from the beginning of the audio	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
utterances.events.word	string	The recognized word or token	Required to calculate silence, overtalk, words per minute, talk rate, and talk streak fields
language	string	Detected or specified language of the audio (a language/locale code)	For a list of supported language codes, see Supported languages.