Swedish Speech Corpus
收藏资源简介:
该数据集是瑞典国家图书馆的KBLab团队与瑞典议会管理、瑞典电视台以及瑞典语言和民间传说研究所合作构建的。数据集包含约56514小时的瑞典语语音数据,涵盖了各种瑞典语的变体,包括方言、社会方言、不同年龄组和口音。数据集来源于多个渠道,如瑞典媒体数据库、瑞典电视台、YouTube、瑞典议会的录音以及方言录音。数据经过预处理,包括音频格式转换、语言检测、转录和筛选,以确保高质量的训练数据。该数据集旨在提升瑞典语音识别系统的性能,使其在低资源语言环境下也能保持竞争力。
This dataset was constructed by the KBLab team at the National Library of Sweden in collaboration with the Swedish Parliament Administration, Sveriges Television, and the Institute for Language and Folklore of Sweden. It contains approximately 56,514 hours of Swedish speech data, covering various Swedish language varieties including regional dialects, sociolects, speech from different age groups and diverse accents. The dataset is sourced from multiple channels, such as the Swedish Media Database, Sveriges Television, YouTube, recordings from the Swedish Parliament, and dialect recordings. The data has undergone preprocessing procedures including audio format conversion, language detection, transcription and filtering to ensure high-quality training data. This dataset aims to enhance the performance of Swedish speech recognition systems, enabling them to remain competitive even in low-resource language environments.




