maxseats/aihub-464-preprocessed-680GB-set-3
收藏Hugging Face2024-06-21 更新2024-06-29 收录
下载链接:
https://hf-mirror.com/datasets/maxseats/aihub-464-preprocessed-680GB-set-3
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含音频、转录文本、输入特征和标签等特征,主要用于语音识别任务。数据集分为训练集、测试集和验证集,分别包含43818、5478和5477个样本。音频文件的采样率为16000Hz,输入特征为log-Mel频谱图,标签为转录文本的token ID。数据集的创建过程涉及音频文件的处理、特征提取、标签生成等步骤,最终将处理后的数据集上传到Hugging Face Hub。
This dataset contains features such as audio, transcripts, input features, and labels, primarily used for speech recognition tasks. The dataset is divided into training, test, and validation sets, containing 43818, 5478, and 5477 samples respectively. The audio files have a sampling rate of 16000Hz, the input features are log-Mel spectrograms, and the labels are token IDs of the transcripts. The dataset creation process involves audio file processing, feature extraction, label generation, and finally uploading the processed dataset to the Hugging Face Hub.
提供机构:
maxseats原始信息汇总
数据集概述
数据集信息
-
特征:
audio: 音频数据,采样率为16000Hz。transcripts: 文本数据,数据类型为字符串。input_features: 序列数据,数据类型为float32。labels: 序列数据,数据类型为int64。
-
分割:
train: 训练集,包含43818个样本,大小为50570714587.90649字节。test: 测试集,包含5478个样本,大小为6322204904.663649字节。valid: 验证集,包含5477个样本,大小为6321050796.429865字节。
-
下载大小: 21857670089字节。
-
数据集总大小: 63213970289.0字节。
配置
- 默认配置:
train: 数据路径为data/train-*。test: 数据路径为data/test-*。valid: 数据路径为data/valid-*。



