Datasets:

Modalities:
Text
Formats:
parquet
Libraries:
Datasets
pandas
License:
yentinglin commited on
Commit
fe3bda7
·
verified ·
1 Parent(s): 26a3c3d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -9
README.md CHANGED
@@ -46,14 +46,21 @@ configs:
46
 
47
  ## Description
48
 
49
- Consolidated from [PeacefulData/HyPoradise-v0](https://huggingface.co/datasets/PeacefulData/HyPoradise-v0)
 
50
 
51
  ## Structure
52
 
53
  ### Data Split
54
 
55
- - **Training Data**: 134,910 entries ~= 2,629,297 for transcript and 13,046,800 for hypothesis (LLama2-tokens)
56
- - **Test Data**: 6,965 entries ~= 131,363 for transcript and 651,099 for hypothesis (LLama2-tokens)
 
 
 
 
 
 
57
 
58
  ### Columns
59
 
@@ -61,23 +68,35 @@ Consolidated from [PeacefulData/HyPoradise-v0](https://huggingface.co/datasets/P
61
  - `transcription`: Corrected asr transcription.
62
  - `hypothesis_concatenated`: An alternative version of the text output.
63
  - `source`: The source of the text entry, indicating the origin dataset.
 
64
  - `score`: An acoustic model score (not all entries have this).
65
 
66
- ### Source Distribution
 
 
67
 
68
  - **Training Sources**:
69
- - `train_cv`: 47,293 entries
70
- - `train_wsj`: 37,514 entries
 
 
 
71
  - `train_swbd`: 36,539 entries
72
  - `train_chime4`: 9,600 entries
73
  - `train_atis`: 3,964 entries
74
-
75
  - **Test Sources**:
 
 
 
76
  - `test_swbd`: 2,000 entries
77
  - `test_cv`: 2,000 entries
78
  - `test_chime4`: 1,320 entries
79
- - `test_wsj`: 836 entries
 
80
  - `test_atis`: 809 entries
 
 
81
 
82
  ## Access
83
 
@@ -90,4 +109,4 @@ dataset = load_dataset("PeacefulData/HyPoradise-pilot")
90
 
91
  ## Acknowledgments
92
 
93
- Thanks https://huggingface.co/datasets/PeacefulData/HyPoradise-v0 for sharing this dataset.
 
46
 
47
  ## Description
48
 
49
+ This dataset is a pilot version of a larger dataset for automatic speech recognition (ASR) corrections across multiple domains.
50
+ It contains paired hypotheses and corrected transcriptions for various ASR tasks consolidated from [PeacefulData/HyPoradise-v0](https://huggingface.co/datasets/PeacefulData/HyPoradise-v0)
51
 
52
  ## Structure
53
 
54
  ### Data Split
55
 
56
+ The dataset is divided into training and test splits:
57
+
58
+ - Training Data: 281,082 entries
59
+ - Approximately 6,255,198 tokens for transcriptions
60
+ - Approximately 31,211,083 tokens for concatenated hypotheses
61
+ - Test Data: 16,108 entries
62
+ - Approximately 327,750 tokens for transcriptions
63
+ - Approximately 1,629,093 tokens for concatenated hypotheses
64
 
65
  ### Columns
66
 
 
68
  - `transcription`: Corrected asr transcription.
69
  - `hypothesis_concatenated`: An alternative version of the text output.
70
  - `source`: The source of the text entry, indicating the origin dataset.
71
+ - `prompt`: Instructional prompt for correction task
72
  - `score`: An acoustic model score (not all entries have this).
73
 
74
+ ### Source Datasets
75
+
76
+ The dataset combines entries from various sources:
77
 
78
  - **Training Sources**:
79
+ - `train_td3`: 50,000 entries
80
+ - `train_other_500`: 50,000 entries
81
+ - `train_cv`: 47,293 entries
82
+ - `train_lrs2`: 42,940 entries
83
+ - `train_wsj_score`: 37,514 entries
84
  - `train_swbd`: 36,539 entries
85
  - `train_chime4`: 9,600 entries
86
  - `train_atis`: 3,964 entries
87
+ - `train_coraal`: 3,232 entries
88
  - **Test Sources**:
89
+ - `test_ls_other`: 2,939 entries
90
+ - `test_ls_clean`: 2,620 entries
91
+ - `test_lrs2`: 2,259 entries
92
  - `test_swbd`: 2,000 entries
93
  - `test_cv`: 2,000 entries
94
  - `test_chime4`: 1,320 entries
95
+ - `test_td3`: 1,155 entries
96
+ - `test_wsj_score`: 836 entries
97
  - `test_atis`: 809 entries
98
+ - `test_coraal`: 170 entries
99
+
100
 
101
  ## Access
102
 
 
109
 
110
  ## Acknowledgments
111
 
112
+ This dataset is consolidated from the PeacefulData/HyPoradise-v0 dataset. Thanks to the original creators for making this data available.