- WER overall and per-speaker
- MTRR on your medical lexicon
- DER and DER-aware WER (DA-WER)
- Per-utterance error analysis
Evaluations
Speech-to-Text Performance Evaluation
Quantifying and visualizing STT results
Medera ships an STT evaluation harness for measuring word error rate (WER), medical term recognition rate (MTRR), and diarization error rate (DER) on your audio. ```bash
medera-eval stt \ —audio./eval/audio/ \ —reference./eval/reference/ \ —report./report.html
The harness produces: