ECD-TSE
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/AI-S2-Lab/EmoCorrector
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为ECD-TSE，是一个专注于情感一致性的基于文本的语音编辑基准数据集，包含了文本变体及其相应情感语音的配对数据。该数据集不仅包含了通过ChatGPT-4和高级TTS系统生成的，传达相似词汇信息但情感不同的合成语音，还涵盖了多位发言人和多样化的情感表达。其任务旨在研究文本基础语音编辑中的情感一致性建模（Tse）。
The dataset named ECD-TSE is a benchmark dataset for text-based speech editing focused on emotional consistency, which contains paired data between text variants and their corresponding emotional speech. This dataset not only includes synthetic speech generated via ChatGPT-4 and state-of-the-art TTS systems that conveys similar lexical information but has distinct emotional tones, but also covers multiple speakers and diverse emotional expressions. The core task of this dataset is to investigate emotional consistency modeling in text-based speech editing (TSE).
提供机构：
Developed by the authors of the paper
