Speech recognition results

Speech recognition results are added to $request.rawRequest. The data is available in the $request.rawRequest.data field.

Example:

{
    "cause": "OK",
    "text": "How are you?"
}

Here:

• cause is the recognition result. Possible values:
  • "OK": the speech was successfully recognized.
  • "NO_INPUT": speech recognition failed.
• text is the speech text. The field is present if cause is "OK".

User request data

The results may also contain additional data about the user’s request in the $request.rawRequest.data.propertiesResponse.result field.

Example:

{
    "recognition_result": {
        "start_time": "0.510s",
        "end_time": "2.550s",
        "gender_identification_result": {
            "male_proba": 0.991876,
            "female_proba": 0.008123994
        }
    },
    "is_final": true
}

User request data is available for the following providers:

• Yandex SpeechKit v2
• Yandex SpeechKit v3
• Tinkoff VoiceKit
• 3iTech

For example, you can get:

• Hypotheses about user gender in Tinkoff Voicekit.
• The list of recognized languages and their probability in Yandex SpeechKit v3.

Supported fields

$request.rawRequest.data.propertiesResponse.result supports a limited number of fields:

Yandex SpeechKit v2

• chunks:

  • alternatives

    • text
    • confidence

  • final

  • endOfUtterance

In the Yandex SpeechKit v2 documentation, you can view the description of these fields and learn about the conditions under which they are added.

Yandex SpeechKit v3

• session_uuid

• audio_cursors

• response_wall_time_ms

• Event (data is added only for Events with the partial or final type):

  • alternatives:

    • words

    • text

    • start_time_ms

    • end_time_ms

    • confidence

    • languages:

      • language_code
      • probability

  • channel_tag

• channel_tag

In the Yandex SpeechKit v3 documentation, you can view the description of these fields and learn about the conditions under which they are added.

Tinkoff VoiceKit

• recognition_result:

  • channel

  • start_time

  • end_time

  • sentiment_analysis_result:

    • negative_prob_audio
    • negative_prob_audio_text

  • gender_identification_result:

    • male_proba
    • female_proba

  • is_final

In the Tinkoff VoiceKit documentation, you can view the description of these fields and learn about the conditions under which they are added.

3iTech

• session_id

• text

• final

• quality_score

• answerphone_detected

• sentiments_info:

  • phrase_sentiments:

    • sentiment_probs
    • phrase_id

  • avg_sentiments_probs

• words_per_second

• age_group

• tech_metrics:

  • data_received_ms

In the 3iTech documentation, you can view the description of these fields and learn about the conditions under which they are added.

info

• Some data is added only if additional settings are specified for the ASR provider. To specify the settings, use the $dialer.setAsrProperty method.
• If you want to use data that is not supported in $request.rawRequest, please contact our support team: support@just-ai.com.