Approximetal commited on
Commit
f5e5cc5
·
verified ·
1 Parent(s): 6d419b7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -0
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:**