pathological_speech / dataset_metadata.json
resproj007's picture
Add dataset metadata JSON
1f1cc4b verified
raw
history blame contribute delete
995 Bytes
{
"description": "Pathological speech evaluation dataset - held-out speakers not used in training/synthesis",
"repo": "resproj007/pathological_speech",
"splits": {
"train": 38741,
"test": 1500
},
"total_samples": 40241,
"test_allocation_by_severity": {
"Mild": 197,
"Moderate": 585,
"Normal": 280,
"Severe": 438
},
"test_by_severity": {
"Mild": 197,
"Moderate": 585,
"Normal": 280,
"Severe": 438
},
"test_by_corpus": {
"UA-Speech": 1069,
"TORGO": 431
},
"test_by_speaker": {
"M01": 179,
"M03": 42,
"M07": 176,
"F04": 167,
"M05": 209,
"F03": 33,
"MC03": 82,
"MC02": 47,
"MC04": 68,
"FC03": 74,
"FC01": 9,
"M10": 193,
"M04": 20,
"M09": 186,
"F01": 15
},
"columns": [
"audio",
"text",
"speaker_id",
"corpus",
"gender",
"condition",
"diagnosis",
"severity",
"intelligibility",
"duration"
],
"audio_sample_rate": 16000
}