Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -86,6 +86,12 @@ Both approaches are valid, but mixing them could lead to evaluation issues.
|
|
86 |
|
87 |
## Audio Processing and Quality Improvements
|
88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
### Silence Trimming
|
90 |
I processed all audio files to remove unnecessary silence and noise:
|
91 |
- Used Silero VAD with a threshold of 0.6 to detect speech segments
|
@@ -201,3 +207,14 @@ The final dataset shows significant improvements:
|
|
201 |
- Maintained original metadata (age, upvotes, etc.)
|
202 |
|
203 |
These improvements make the dataset more suitable for training speech recognition models while maintaining the diversity and richness of the original CommonVoice collection.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
|
87 |
## Audio Processing and Quality Improvements
|
88 |
|
89 |
+
### Audio Resampling
|
90 |
+
All audio files were resampled to 16 kHz to:
|
91 |
+
- Make the dataset directly compatible with Whisper and similar models
|
92 |
+
- Eliminate the need for runtime resampling during training
|
93 |
+
- Ensure consistent audio quality across the dataset
|
94 |
+
|
95 |
### Silence Trimming
|
96 |
I processed all audio files to remove unnecessary silence and noise:
|
97 |
- Used Silero VAD with a threshold of 0.6 to detect speech segments
|
|
|
207 |
- Maintained original metadata (age, upvotes, etc.)
|
208 |
|
209 |
These improvements make the dataset more suitable for training speech recognition models while maintaining the diversity and richness of the original CommonVoice collection.
|
210 |
+
|
211 |
+
## Tools Used
|
212 |
+
|
213 |
+
This dataset processing work was completed using [ASRTK (Automatic Speech Recognition Toolkit)](https://github.com/ysdede/asrtk), an open-source Python toolkit designed to streamline the development and enhancement of ASR systems. ASRTK provides utilities for:
|
214 |
+
|
215 |
+
- Audio processing with advanced splitting and resampling capabilities
|
216 |
+
- Text normalization and cleaning
|
217 |
+
- Forced alignment using Silero VAD models
|
218 |
+
- Efficient batch processing with multi-threading support
|
219 |
+
|
220 |
+
The toolkit is available under the MIT license and welcomes contributions from the community.
|