ParlaSpeech
收藏资源简介:
ParlaSpeech数据集是由斯洛文尼亚约瑟夫·斯特凡研究所创建的,基于欧洲26个国家的议会记录和录音的大型语音与文本对齐数据集。该数据集包含超过5000小时的语音和相应的文本转录,主要涵盖克罗地亚语、波兰语和塞尔维亚语。数据集的创建过程涉及复杂的文本与音频对齐技术,旨在解决多语言环境下语音数据的稀缺问题。该数据集的应用领域广泛,包括自动语音识别、语言学研究和政治科学研究等。
ParlaSpeech is a large-scale aligned speech-text dataset developed by the Jožef Stefan Institute in Slovenia, which is constructed based on parliamentary records and audio recordings from 26 European countries. It contains more than 5,000 hours of speech paired with corresponding text transcriptions, primarily covering Croatian, Polish and Serbian languages. The dataset development involves sophisticated text-audio alignment technologies, aiming to address the scarcity of speech data in multilingual scenarios. It has a wide range of application fields including automatic speech recognition, linguistic research and political science research.




