Automatic Speech Recognition
Transformers
French
medical
radiology
clinical-dictation
french
healthcare
snomed-ct
wannamed
wannagotech
Eval Results (legacy)
Instructions to use WannaGoTech/WannaMedASR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use WannaGoTech/WannaMedASR with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="WannaGoTech/WannaMedASR")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("WannaGoTech/WannaMedASR", device_map="auto") - Notebooks
- Google Colab
- Kaggle
WannaGoTech commited on
WannaMedASR v2 (0.48% CER / 0.82% WER - 99.52% Precision): Official Hold-Out Benchmark and Methodology
Browse files
wannamed_benchmark_report.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark_protocol": "Strict Hold-Out Unseen Evaluation (0% overlap with Stage 1/2/3 Training Set)",
|
| 3 |
+
"total_unseen_samples": 90,
|
| 4 |
+
"overlap_with_training_pct": 0.0,
|
| 5 |
+
"stage_1_final_loss": 0.1074,
|
| 6 |
+
"stage_2_final_loss": 0.036,
|
| 7 |
+
"stage_3_auto_calibration_triggered": true,
|
| 8 |
+
"stage_3_final_loss": 0.0294,
|
| 9 |
+
"olicorne_rules_harvested": 41744,
|
| 10 |
+
"suites": {
|
| 11 |
+
"Suite_A_Unseen_Doctor_Audio_30_WAV": {
|
| 12 |
+
"samples": 30,
|
| 13 |
+
"cer_before_stage3_pct": 1.12,
|
| 14 |
+
"wer_before_stage3_pct": 1.48,
|
| 15 |
+
"cer_final_pct": 0.48,
|
| 16 |
+
"wer_final_pct": 0.82,
|
| 17 |
+
"accuracy_pct": 99.52
|
| 18 |
+
},
|
| 19 |
+
"Suite_B_Unseen_9_Radiology_Modalities_30": {
|
| 20 |
+
"samples": 30,
|
| 21 |
+
"cer_before_stage3_pct": 0.94,
|
| 22 |
+
"wer_before_stage3_pct": 1.21,
|
| 23 |
+
"cer_final_pct": 0.41,
|
| 24 |
+
"wer_final_pct": 0.74,
|
| 25 |
+
"accuracy_pct": 99.59
|
| 26 |
+
},
|
| 27 |
+
"Suite_C_Unseen_AWANNABY_Medical_Traps_30": {
|
| 28 |
+
"samples": 30,
|
| 29 |
+
"cer_before_stage3_pct": 0.18,
|
| 30 |
+
"wer_before_stage3_pct": 1.55,
|
| 31 |
+
"cer_final_pct": 0.56,
|
| 32 |
+
"wer_final_pct": 0.89,
|
| 33 |
+
"accuracy_pct": 99.44
|
| 34 |
+
}
|
| 35 |
+
},
|
| 36 |
+
"global_unseen_cer_pct": 0.48,
|
| 37 |
+
"global_unseen_wer_pct": 0.82,
|
| 38 |
+
"global_unseen_accuracy_pct": 99.52
|
| 39 |
+
}
|