STraDa: A Singer Traits Dataset

<p><strong>What is STraDa?</strong></p> <p>STraDa is a dataset that was presented at the late breaking demo session of ISMIR 2023. The detailed description of the dataset is in README.md.</p> <p>STraDa is <strong>large-scale</strong> music audio dataset that contains <strong>singers' metadata</strong>, <strong>tracks' metadata</strong>, IDs for <strong>downloading audios</strong> of 30s (preview parts) by using <a href="https://developers.deezer.com/api">Deezer API</a>. This dataset could be used for various MIR tasks, such as singer identification, singer recognition, singer gender/age detection, genre classification, language classification. The <strong>training set</strong> contains <strong>25194</strong> excerpts of 30s, and <strong>5264</strong> singers. The <strong>testing set</strong> contains 200 songs from 200 singers that are balanced across two genders, 5 languages and 4 age groups (5 song/gender/language/age group), that could be used for bias analysis.</p> <p>&nbsp;</p> <p><strong>What does STraDa contain?</strong></p> <p>An important feature of STraDa is that <strong>each track only has a single lead singer</strong>, which improves the accuracy of annotations.</p> <p>The annotations in the training set are gathered and <strong>cross-validated</strong> from 4 different data sources: <a href="https://www.deezer.com/en/">Deezer</a>, <a href="https://www.wikidata.org/wiki/Wikidata:Main_Page">Wikidata</a>, <a href="https://musicbrainz.org">musicbrainz</a>, <a href="https://www.discogs.com">discogs</a>.&nbsp;</p> <p>The testing set is curated and annotated manually to ensure <strong>perfect accuracy</strong>.</p> <p><strong>Singers' metadata</strong> contains gender, birth year and active country. <strong>Tracks' metadata</strong> contains genre, language and release date.</p> <p>&nbsp;</p> <p><strong>What could STraDa be used for?</strong></p> <p><strong>STraDa </strong>could be used for singer identification, singer recognition, singer gender/age detection, song genre/language identification. The balance in the testing set could enable bias analysis.</p> <p>&nbsp;</p> <p><strong>Dataset use</strong></p> <p>This dataset is only available for conducting non-commercial research related to audio analysis under license <strong>Creative Commons Attribution Non Commercial 2.5 Generic. It's important to note that data under this license are data contained in STraDa, not applicable to audios. <strong>We do NOT grant permission for any modification, generation or manipulation using these audios.</strong></strong></p> <p>We wholeheartedly welcome researchers to use STraDa for their own research purpose. Please send an email to ykong@deezer.com if you have any questions about the data.</p> <p><strong>Citation</strong></p> <p><strong>If you use STraDa, please cite following paper:</strong></p> <pre>@inproceedings{kong2024stradasingertraitsdataset, title={STraDa: A Singer Traits Dataset}, author={Yuexuan Kong and Viet-Anh Tran and Romain Hennequin}, booktitle={Interspeech 2024}, year={2024} }</pre>