SttOverSuppressionScorer
Pairwise over-suppression detector. Two physics-grounded signals: musical noise (per-bin temporal-kurtosis ratio of cleaned vs raw in the speech band; Esch & Vary, ICASSP 2009) and flattened modulation (loss of 4-16 Hz syllable-rate envelope energy). No LLM, no GPU — numpy + librosa + soundfile, ~30 ms per pair.
Overview
Pairwise over-suppression detector. Two physics-grounded signals: musical noise (per-bin temporal-kurtosis ratio of cleaned vs raw in the speech band; Esch & Vary, ICASSP 2009) and flattened modulation (loss of 4-16 Hz syllable-rate envelope energy). No LLM, no GPU — numpy + librosa + soundfile, ~30 ms per pair.
Use Cases
- General AI evaluation and quality assessment
How It Works
This scorer uses deterministic rule-based evaluation to validate outputs against specific criteria. It applies predefined rules and patterns to assess the response, providing consistent and reproducible results without requiring LLM inference.
Input Schema
| Parameter | Type | Required | Description |
|---|---|---|---|
| stt_data.metadata | any | Yes | - |
| stt_data.audio_uuid (or stt_data.audio_url) | any | Yes | - |
| stt_data.raw_audio_uuid (or stt_data.raw_audio_url) | any | Yes | - |
Output Schema
| Field | Type | Description |
|---|---|---|
| score | any | - |
| passed | any | - |
| reasoning | any | - |
| metadata | any | - |
Score Interpretation
Related Scorers
SttEfficacyScorer
Scores transcript fidelity vs spoken audio (1-10). Parses JSON score and reasoning; metadata may inc...
Voice & AudioAgentWpmScorer
Agent speaking rate (words per minute) from assistant turn text and VAD-gated agent speech time. con...
Voice & AudioAiInterruptUserScorer
Counts agent barge-in events (agent starts speaking while user utterance is open). Zero events → 10....
Voice & AudioAssistantAveragePitchScorer
Mean fundamental frequency (Hz) on the assistant channel using librosa yin, mapped to 1-10 for thres...
Frequently Asked Questions
When should I use this scorer?
Use SttOverSuppressionScorer when you need to evaluate audio aspects of your AI outputs. It's particularly useful for general ai evaluation and quality assessment.
Why doesn't this scorer need expected output?
This scorer evaluates quality aspects that don't require comparison against a reference answer. It uses the system prompt and context as the implicit ground truth.
Can I customize the threshold?
Yes, the default threshold of 7 can be customized when configuring the scorer.
Quick Info
Ready to try SttOverSuppressionScorer?
Start evaluating your AI agents with Noveum.ai's comprehensive scorer library.
Explore More Scorers
Discover 112 calibrated LLM-as-Judge scorers for comprehensive AI evaluation