ATT-Corpus
收藏arXiv2025-09-30 收录
数据链接:
官方服务:
资源简介:
该数据集是一个多维度的中文语料库,旨在通过一个受图灵测试启发的简化评估协议,用于评估文本转语音(TTS)系统的人性化程度。参与者为他们的评估提供了书面理由,这为定性分析提供了支持。该数据集旨在减少评分偏差,提高评估的稳健性。它包含了20种不同的语音风格,横跨5个TTS模型家族,并由857名参与者进行了评估,其任务是评估TTS系统的人性化程度。
This dataset is a multi-dimensional Chinese corpus designed to evaluate the human likeness of text-to-speech (TTS) systems via a simplified evaluation protocol inspired by the Turing Test. Participants provided written justifications for their evaluations, which supports qualitative analysis. This dataset aims to reduce scoring bias and enhance the robustness of evaluations. It encompasses 20 distinct speech styles spanning 5 TTS model families, and was evaluated by 857 participants whose assigned task was to assess the human likeness of the TTS systems.
搜集汇总
数据集介绍

以上内容由遇见数据集搜集并总结生成



