Language Models

This documentation provides a comprehensive overview of the language models available in the Speech Recognition (SR) service and how to select the appropriate version for your needs.

Endpoint Details

• URL: {{Address}}/v1/speech/dictation/models
• Actor: REST Service Client
• Goal: Retrieve the list of available language models.
• Pre-Conditions: None

Example API Response

When querying the available language models, you will receive a response similar to the following:

{
    "Identifier": {
        "Tenant": "Default",
        "Name": "English",
        "Variant": 4,
        "Version": 1
    }
}

Understanding Model Versions

The Model Version parameter allows you to select a specific model version for processing your submitted audio.

Using the Latest Model Version

To use the latest version of a model:

• Set ModelVersion=0
• Or simply omit the ModelVersion parameter.

📌 Note: Whether you explicitly set ModelVersion=0 or do not include the parameter at all, the system will automatically select the most up-to-date and optimized language model, ensuring you benefit from the latest advancements in recognition accuracy and performance.

Using a Specific Model Version

If you need to use an older version of a language model:

1. Retrieve the available versions by making a GET request to the models endpoint.
2. Use the ModelVersion parameter to specify the desired version when making a transcription request.

📌 Mapping Note:
The Variant field in the model response corresponds to the ModelVersion parameter when making a request.

Example Request for a Specific Model Version

curl --location '{{Address}}/v1/speech/dictation/request' \
--header 'Content-Type: audio/wave' \
--header 'ModelName: Turkish' \
--header 'ModelVersion: 4' \
--header 'Tenant: Default' \
--header 'Authorization: Bearer Token' \
--data 'audio.wav'

When to Use Model Versions?

• ✅ When you want to use the latest version, no additional configuration is required.
• ✅ When you need an earlier version for consistency, specify the desired ModelVersion in your request.