The data can be downloaded from https://www.kaggle.com/toponowicz/spoken-language-identification. This dataset contains FLAC audio files and the task is to identify the language for each file. More dataset descriptions and the open-source models can be found at https://github.com/tomasz-oponowicz/spoken_language_identification.

Note the downloaded files should be organized by the language types, i.e., the audio files with the same language must be put under a same directory. This directory will be applied to generate the labels for each file.