rendchevi/emilia-yodas-english-neucodec-VJKL-250k-prep

这是一个多模态语音-文本数据集，包含语音和对应的文本信息，适用于语音合成、语音质量评估或语音识别等任务。数据集特征包括语音质量评分（dnsmos）、持续时间（duration）、语言（language）、说话人数量（phone_count）、说话人标识（speaker）、文本内容（text）、编码序列（codes）、音素序列（phoneme）和说话人嵌入向量（speaker_embedding）。数据集分为训练集（25,000个样本）、评估集（2,500个样本）和测试集（25,000个样本），总大小约为317 MB。

This is a multimodal speech-text dataset containing speech and corresponding text information, suitable for tasks such as speech synthesis, speech quality assessment, or speech recognition. The dataset features include speech quality score (dnsmos), duration, language, phone count, speaker identifier, text content, code sequences, phoneme sequences, and speaker embedding vectors. It is divided into training set (25,000 samples), evaluation set (2,500 samples), and test set (25,000 samples), with a total size of approximately 317 MB.