Confidence scores

Overview

Soniox Speech-to-Text AI provides a confidence score for each recognized token (word or sub-word) in the transcript. The confidence score represents the AI model's estimate of how likely the token was recognized correctly.

Confidence values range from 0.0 to 1.0, where:

• 1.0 indicates very high confidence
• 0.0 indicates very low confidence

Lower confidence scores may indicate that the token was difficult to recognize, possibly due to background noise, accents, unclear speech, or uncommon vocabulary.

Confidence scores can be used to assess transcription quality, highlight uncertain words, or post-process results for improved accuracy.

Output format

Each token in the API response includes:

• text: The recognized token (word or sub-word)
• confidence: The confidence score for the token

Example response

In this example, the word "beautiful" is split into three tokens, each with an associated confidence score:

{
  "tokens": [
    {"text": "Beau", "confidence": 0.82},
    {"text": "ti",  "confidence": 0.87},
    {"text": "ful", "confidence": 0.98}
  ]
}

Use cases