Supported languages and recognition models

A recognition model is a model trained to recognize speech in a particular language. The models are trained on datasets generated by Yandex services and applications. This allows us to keep improving the speech recognition quality.

The general model is the main supported model for each recognition type. It recognizes speech on any topic in a given language, including short and long utterances, names, addresses, dates, and numbers.

Version tagsVersion tags

Three versions of the general model may be available at the same time. Select the one you need by tag:

• general: Main version.
• general:rc: Release candidate version available for testing.
• general:deprecated: Previous version.

You can also use the deferred-general tag for asynchronous recognition. Learn more about asynchronous recognition modes.

Supported recognition languagesSupported recognition languages

Use a recognition language code from the table below. All available code values are case insensitive.

Automatic language detectionAutomatic language detection

SpeechKit automatically detects the language in each sentence during speech recognition.

To use automatic language detection, specify auto in the language_code parameter of the LanguageRestrictionOptions() method:

language_restriction=stt_pb2.LanguageRestrictionOptions(
              restriction_type=stt_pb2.LanguageRestrictionOptions.WHITELIST,
              language_code=['auto']
        )
        

Along with recognition results, the service returns language labels containing the language code and probability of its correct detection:

language_code: "ru-RU" probability: 0.91582357883453369
        

If a sentence contains words in different languages, the language may be detected incorrectly. To make recognition more accurate, replace auto with a list of expected languages as a clue for the model. For example:

...
              language_code=['en-US', 'es-ES', 'fr-FR']
        ...
        

The language is detected for each phrase. If a sentence has phrases in different languages, all of them will most likely be transcribed in the same language.

Examples

Recognition accuracyRecognition accuracy

You can assess the recognition quality of a SpeechKit model yourself using your own data, e.g., in Yandex DataSphere.

Use casesUse cases

• Streaming speech recognition with auto language detection using the API v3
• Audio file streaming recognition using the API v3
• Microphone speech streaming recognition using the API v3

See alsoSee also

• Supported audio formats
• Streaming speech recognition
• Synchronous audio recognition
• Asynchronous recognition
• Extending a speech recognition model
• Streaming speech recognition with auto language detection using the API v3