UdS-LSV/Saar-Voice
收藏资源简介:
Saar-Voice是一个多说话人语音语料库,专注于德国萨尔布吕肯及周边地区使用的莱茵法兰克方言(通常称为Saarländisch)。数据集包含9位说话人(P01-P09),总时长约6小时,录音句子4,871个,未录音句子3,901个。音频采样率为22,050 Hz,格式为WAV(立体声,16位)。数据分为训练集(2,373句)、测试集(1,510句)、验证集(268句)和保留集(720句,用于未见说话人评估)。文本来源包括MASSIVE数据集的德语子集(本地化为Saarländisch方言正字法)、书籍、诗歌、短篇小说集以及当地作者社区提供的文本。数据集旨在支持方言语音处理任务,如语音合成和识别。
Saar-Voice is a multi-speaker speech corpus for the Rhine Franconian dialect of German as spoken in Saarbrücken and the surrounding region, loosely referred to as Saarländisch. The dataset includes 9 speakers (P01–P09), with a total duration of approximately 6 hours, 4,871 recorded sentences, and 3,901 unrecorded sentences. Audio is sampled at 22,050 Hz in WAV format (stereo, 16-bit). Splits include train (2,373 sentences), test (1,510 sentences), validation (268 sentences), and held_out (720 sentences, for unseen speaker evaluation). Text sources include the German subset of the MASSIVE dataset (localized into Saarländisch dialect orthography), books, poems, short story collections, and texts provided by the local author community. The dataset is designed to support dialect speech processing tasks such as speech synthesis and recognition.




