Update README.md
Browse files
README.md
CHANGED
|
@@ -48,6 +48,7 @@ It contains a large-scale training set (150k+ hours) and a curated evaluation se
|
|
| 48 |
|
| 49 |
### Train Set
|
| 50 |
|
|
|
|
| 51 |
The training set is constructed by filtering large-scale aligned audio–text pairs with language- and dataset-specific constraints.
|
| 52 |
|
| 53 |
**Filtering rules:**
|
|
@@ -76,6 +77,7 @@ Words and chars statistics are computed based on the normalized alignment text (
|
|
| 76 |
|
| 77 |
### Eval Set
|
| 78 |
|
|
|
|
| 79 |
The eval set is built by filtering, trimming, and ranking aligned audio–text pairs.
|
| 80 |
|
| 81 |
**Filtering rules:**
|
|
|
|
| 48 |
|
| 49 |
### Train Set
|
| 50 |
|
| 51 |
+
https://huggingface.co/datasets/LEMAS-Project/LEMAS-Dataset-train
|
| 52 |
The training set is constructed by filtering large-scale aligned audio–text pairs with language- and dataset-specific constraints.
|
| 53 |
|
| 54 |
**Filtering rules:**
|
|
|
|
| 77 |
|
| 78 |
### Eval Set
|
| 79 |
|
| 80 |
+
https://huggingface.co/datasets/LEMAS-Project/LEMAS-Dataset-eval
|
| 81 |
The eval set is built by filtering, trimming, and ranking aligned audio–text pairs.
|
| 82 |
|
| 83 |
**Filtering rules:**
|