You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

ScreenTalk_ZH2ZH

🎨 ScreenTalk_ZH2ZH 是一个中文电影和电视剧对话数据集,包含高质量的 中文语音转写对(ZH → ZH),适用于自动语音识别(ASR)模型的训练与微调。


📦 数据概览

  • 语言:中文(简体)

  • 任务类型:自动语音识别(ASR)

  • 样本格式

    • audio:本地音频文件路径(.wav 格式)
    • sentence:与音频对应的中文转写文本

示例:

{
  "audio": "./audio/zh-CN/20250627161128/clip_b00fca74-e8d5-4e0e-8a04-731de52e4f22.wav",
  "sentence": "我马上就来,别着急。"
}

🧪 用例推荐

  • 微调 Whisper 模型以提升中文识别性能
  • 用作中文语音识别测试集、验证集
  • 语音对齐、自动字幕生成项目
  • 中文口话语料构建与标注辅助

🔧 加载与使用(🧯 Datasets)

from datasets import load_dataset

dataset = load_dataset("Itbanque/ScreenTalk_ZH2ZH", split="train", use_auth_token=True)

# 查看样本
print(dataset[0])

🧠 数据集特点

  • 来源于真实影视作品,语气自然,场景丰富
  • 语音切片长度控制在数秒以内,便于模型训练
  • 所有音频均为 .wav 格式,16kHz 采样率

📜 许可与使用条款

  • 本数据集仅供研究、教育用途。
  • 禁止用于任何形式的商业用途。
  • 使用或引用本数据集时,请注明:
Dataset by Itbanque. ScreenTalk_ZH2ZH. 2025. https://huggingface.co/datasets/Itbanque/ScreenTalk_ZH2ZH

📬 联系方式

若有任何问题或合作意向,请访问:https://www.itbanque.com


© 2025 Itbanque. All rights reserved.

Downloads last month
0