Transcript parameters

Identifying speakers

To identify speakers on the recording, you can either:

• use the multichannel parameter and Sinch Contact Pro's Record in Stereo option (System Services > Recording Settings), which will label the speakers as Agent and Customer

  This will only identify one speaker per channel so for example consultation calls will show both Sinch Contact Pro users as Agent.

  This option is in use by default.

• use the speaker_labels parameter, which will label the speakers as Speaker:1, Speaker:2 and so on

  "transcript_parameters": {
          "speaker_labels": true
        }

Language handling

You can either define the language for the transcription or let the AI detect the language. By default, the system uses language detection which will try to identify the language spoken in the recording. However, if you want to define the language yourself, you can do it with the language_code parameter.

"transcript_parameters": {
        "language_code": "fi"
      }

You can see the supported languages on the Overview page.

Sentiment analysis

By default, a sentiment analysis (Positive/Negative/Neutral) is done to the call recording. If you don't want the analysis to be done, define parameter sentiment_analysis as false.

"transcript_parameters": {
        "sentiment_analysis": false
      }

Removing personal identifiable information (PII)

It's possible to automatically identify and remove sensitive information from the transcript. By default, the system removes banking information, credit card numbers and CVV, and passport numbers. To override the defaults and configure the removed data yourself, use the redact_pii_policies parameter to list the data to remove:

"transcript_parameters": {
        "redact_pii_policies": [
          "credit_card_number",
          "credit_card_cvv",
          "banking_information",
          "person_name"
        ]
      }

To disable PII redaction, define parameter redact_pii as false:

"transcript_parameters": {
        "redact_pii": false
      }

Example of full JSON configuration with transcript parameters

{
  "request": {    
    "method": "POST",
    "authentication": {
      "type": "oauth",
      "secrets_ref": "orchestration-layer"
    },
    "headers": {
      "x-cmi-api-key": "{secrets.ri-apikey}",
      "x-cmi-password": "{secrets.cmi-user}"
    },
    "body": {
      "s3_key": "{contact.cad.s3_key}",
      "s3_bucket": "{contact.cad.s3_bucket}",
      "s3_root": "{contact.cad.s3_root}",
      "cmi_url": "https://8h6b51di7x.execute-api.eu-central-1.amazonaws.com/acme/cmi",
      "cmi_user": "transcript.user",
      "ai_analytics": true,
      "recording_id": "{contact.cad.recording_id}",
      "contact_id": "{contact.id}",
      "s3_region": "{contact.cad.s3_region}",
      "transcript_parameters": {
        "language_code": "fi",
        "multichannel": true,
        "sentiment_analysis": false,
        "redact_pii": true,
        "redact_pii_policies": [
          "credit_card_number",
          "credit_card_cvv",
          "banking_information",
          "person_name"
        ]
      }
    },
    "timeout": 60
  }
}

If you also want to do an AI analysis on the conversation, see AI analysis.