Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: [hau]
|
| 3 |
+
tags:
|
| 4 |
+
- asr
|
| 5 |
+
- speech-correction
|
| 6 |
+
- mt5
|
| 7 |
+
- multilingual
|
| 8 |
+
license: apache-2.0
|
| 9 |
+
datasets: []
|
| 10 |
+
base_model: google/mt5-base
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# MT5-based ASR Corrector for `hau`
|
| 14 |
+
|
| 15 |
+
This model is fine-tuned on MT5-base to correct ASR outputs for `hau`.
|
| 16 |
+
It was trained using pairs of reference scripts and ASR outputs generated by a fine-tuned Whisper Large model.
|
| 17 |
+
|
| 18 |
+
### Use Case
|
| 19 |
+
Given a noisy ASR output, this model outputs a corrected transcription closer to the reference.
|
| 20 |
+
|
| 21 |
+
### Base Model
|
| 22 |
+
[google/mt5-base](https://huggingface.co/google/mt5-base)
|
| 23 |
+
|
| 24 |
+
### Languages
|
| 25 |
+
This model is trained specifically for: **hau**
|