kotoba-tech/kotoba-whisper-eval
收藏Hugging Face2024-04-21 更新2025-04-08 收录
下载链接：
https://hf-mirror.com/datasets/kotoba-tech/kotoba-whisper-eval
下载链接
链接失效反馈官方服务：
资源简介：
---
dataset_info:
features:
- name: audio
dtype: audio
splits:
- name: train
num_bytes: 109272060.0
num_examples: 4
download_size: 106357055
dataset_size: 109272060.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---
# Audio Only Eval Dataset
This repo contains a few audio examples for kotoba-whisper evaluation (no transcription). Currently, it includes 1-hour presentation audio, and three manzai recordings (five min each).
| path | array size | sampling rate | duration (min) |
|:---------------------|----------:|----------------:|---------:|
| long_interview_1.mp3 | 133,197,167 | 44,100 | 50.3 |
| manzai3.mp3 | 5,343,599 | 16,000 | 5.6 |
| manzai2.mp3 | 4,703,434 | 16,000 | 4.9 |
| manzai1.mp3 | 5,383,343 | 16,000 | 5.6 |
提供机构：
kotoba-tech
