sage-bergerson/edacc_whisper

该数据集包含代码、音频、转录文本和口音四个主要特征。数据集仅包含一个训练集，共有19137个样本，总大小为5883385729.599字节。数据集的下载大小为6949751651字节。配置部分指定了数据文件的路径。

The dataset is designed for fine-tuning the Whisper model, based on the Edinburgh International Accents of English Corpus. It includes segmented and processed audio along with corresponding transcripts. The data is divided into FT_data for fine-tuning and EVAL_data for testing, allowing users to decide their own train/test split. FT_data contains tokenized transcripts and log-mel spectrogram input features, while EVAL_data includes audio files and corresponding transcripts. Additionally, there is an accents.csv file containing accent information corresponding to each segment.

原始信息汇总

数据集概述

数据集特征

• code: 字符串类型
• audio: 音频类型
• transcript: 字符串类型
• accent: 字符串类型

数据集分割

• train:
  • 样本数量: 19137
  • 数据大小: 5883385729.599 字节

数据集大小

• 下载大小: 6949751651 字节
• 数据集大小: 5883385729.599 字节

配置

• default:
  • 数据文件路径: data/train-*