Sensitive data recognition

Find and label personal, health, and payment card information in Batch transcripts.

Sensitive data recognition finds sensitive data in a Batch transcript, such as names, dates of birth, health conditions, and credit card numbers. Each value is returned in the JSON transcript as an entity result labeled with its entity class.

Sensitive data recognition is available for Batch transcription in Speechmatics SaaS, on all models. It is not available in Realtime or on-prem deployments.

Enable sensitive data recognition

To enable sensitive data recognition, set enable_entities to true and entities_config.detect_advanced_entities to true:

{
  "type": "transcription",
  "transcription_config": {
    "language": "en",
    "enable_entities": true,
    "entities_config": {
      "detect_advanced_entities": true
    }
  }
}

entities_config is only accepted when enable_entities is true.

Sensitive data recognition output

For the utterance "Mark was born on the 25th of July 2000.", the JSON transcript looks like this (timings omitted):

{
  "results": [
    {
      "type": "entity",
      "entity_class": "name_given",
      "alternatives": [{ "content": "Mark", "speaker": "S1" }]
    },
    { "type": "word", "alternatives": [{ "content": "was", "speaker": "S1" }] },
    { "type": "word", "alternatives": [{ "content": "born", "speaker": "S1" }] },
    { "type": "word", "alternatives": [{ "content": "on", "speaker": "S1" }] },
    { "type": "word", "alternatives": [{ "content": "the", "speaker": "S1" }] },
    {
      "type": "entity",
      "entity_class": "dob",
      "alternatives": [{ "content": "25th of July 2000", "speaker": "S1" }]
    },
    { "type": "punctuation", "alternatives": [{ "content": ".", "speaker": "S1" }] }
  ]
}

Each recognized value has type set to entity and an entity_class, such as name_given or dob. Recognition returns the most specific class that applies. For the full list, see Sensitive entity classes.

To remove the values instead of labeling them, use redaction. You can enable both in the same job.

Sensitive data recognition and smart formatting entities

enable_entities returns one of two kinds of entities:

	Smart formatting entities	Sensitive data recognition
Configuration	enable_entities: true	enable_entities: true and entities_config.detect_advanced_entities: true
Entity classes	cardinal, date, money, percentage, and other smart formatting classes	Sensitive entity classes, such as name_given, dob, and credit_card
spoken_form and written_form	Included	Not included
Models	Standard and Enhanced	All models

A job returns one kind or the other, not both. On Melia 1 and Oak 1, enable_entities is only accepted with detect_advanced_entities set to true.

Supported languages for sensitive data recognition

Sensitive data recognition supports all transcription languages, except:

• Bashkir
• Basque
• Esperanto
• Galician
• Interlingua
• Irish
• Maltese
• Marathi
• Mongolian
• Urdu
• Uyghur
• Welsh

For Mandarin, use the Simplified output locale (cmn-Hans, the default). Sensitive data recognition of Traditional characters (cmn-Hant) isn't supported yet.

Next steps

• Remove sensitive data with redaction.
• Look up every class in Sensitive entity classes.
• See all parameters in the Batch API reference.