Smallest ASR Labelling Framework — pass ≥ 97%

Voice Annotation Scorer

Paste or drop an annotator JSON and the gold reference JSON. The scorer compares them per the framework rulebook and returns a numeric score with a full breakdown of every HARD violation and SOFT disagreement.

Building a tool? See the public API docs  ·  openapi.json

Annotator JSON

Gold JSON

Threshold: 0.97 — client SLA per framework target (WER < 3%)

How buckets roll up to the overall score

Scoring math