BSD35k-CS (Broad Sound Dataset 35k - Crowd Sourced)

The BSD35k-CS dataset (Broad Sound Dataset 35k - Crowd Sourced) is an open collection of human-labeled sounds containing about 35k Freesound audio clips, annotated according to the second-level classes defined in the Broad Sound Taxonomy (BST). BST is currently being used in Freesound for organization, filtering, and post-processing tasks. The dataset was created at the Music Technology Group of Universitat Pompeu Fabra. This dataset is a complement to the already existing BSD10k dataset, and triples the size of it. However, unlike BSD10k in which sounds were annotated by experts, the annotations of BSD35k-CS come from the authors of the sounds themselves, who uploaded the sounds to Freesound and selected a BST category for them. That is why we call it Crowd Sourced. It should therefore be expected that the annotations in BSD35k-CS are more noisy and less coherent than in BSD10k. All the sounds included in BSD35k-CS were uploaded in Freesound between April 1st 2025 and Janurary 27th 2026 (both included).

Citation

When using all or part of the BSD35k-CS dataset, please cite the corresponding Zenodo page: 

@dataset{anastasopoulou_2026_19187100,
  author = {Anastasopoulou, Panagiota and Font Corbera, Frederic},
  title = {BSD35k-CS (Broad Sound Dataset 35k - Crowd Sourced)},
  month = mar,
  year = 2026,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.19187100},
  url = {https://doi.org/10.5281/zenodo.19187100},
}

License

BSD35k-CS is released in its entirety under the CC BY 4.0 license. We note, though, that each audio file is released under its own Creative Commons (CC) license, as defined by the respective uploader in Freesound. Some sounds require attribution to their original authors, while others forbid commercial reuse. If the dataset is used in a commercial setting, the sounds with CC BY-NC licenses should be excluded.

This is the distribution of sounds per license:

• CC0: 25,035
• CC BY: 6,811
• CC BY-NC: 1,983

Links to the license deeds for each sound can be further accessed through BSD350k-CS_metadata.csv.

Data structure

BSD35k-CS can be accessed as follows:

𝚛𝚘𝚘𝚝/
├── 𝚊𝚞𝚍𝚒𝚘/                                                           𝙰𝚞𝚍𝚒𝚘 𝚏𝚒𝚕𝚎𝚜
├── 𝚖𝚎𝚝𝚊𝚍𝚊𝚝𝚊/                                                     𝙼𝚎𝚝𝚊𝚍𝚊𝚝𝚊 𝚏𝚒𝚕𝚎𝚜
 │   ├── 𝙱𝚂𝙳35k-CS_𝚖𝚎𝚝𝚊𝚍𝚊𝚝𝚊.𝚌𝚜𝚟                    𝙳𝚊𝚝𝚊𝚜𝚎𝚝'𝚜 𝚖𝚎𝚝𝚊𝚍𝚊𝚝𝚊
 │   ├── 𝙱𝚂𝚃_𝚍𝚎𝚜𝚌𝚛𝚒𝚙𝚝𝚒𝚘𝚗.𝚌𝚜𝚟                           𝚃𝚊𝚡𝚘𝚗𝚘𝚖𝚢 𝚒𝚗𝚏𝚘𝚛𝚖𝚊𝚝𝚒𝚘𝚗
 │   └── 𝙱𝚂𝚃_𝚍𝚒𝚊𝚐𝚛𝚊𝚖.𝚙𝚗𝚐                                   𝚃𝚊𝚡𝚘𝚗𝚘𝚖𝚢 𝚍𝚒𝚊𝚐𝚛𝚊𝚖
├── 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚜/                                                     𝙿𝚛𝚎𝚌𝚘𝚖𝚙𝚞𝚝𝚎𝚍 𝚎𝚖𝚋𝚎𝚍𝚍𝚒𝚗𝚐𝚜
 │   ├── 𝚌𝚕𝚊𝚙_𝚊𝚞𝚍𝚒𝚘_𝚎𝚖𝚋𝚎𝚍𝚍𝚒𝚗𝚐𝚜                       𝙰𝚞𝚍𝚒𝚘 𝚎𝚖𝚋𝚎𝚍𝚍𝚒𝚗𝚐𝚜
 │   └── 𝚌𝚕𝚊𝚙_𝚝𝚎𝚡𝚝_𝚎𝚖𝚋𝚎𝚍𝚍𝚒𝚗𝚐𝚜                         𝚃𝚎𝚡𝚝 𝚎𝚖𝚋𝚎𝚍𝚍𝚒𝚗𝚐𝚜
└── 𝚁𝙴𝙰𝙳𝙼𝙴.𝚖𝚍                                                    𝙳𝚘𝚌𝚞𝚖𝚎𝚗𝚝𝚊𝚝𝚒𝚘𝚗 (𝚝𝚑𝚊𝚝 𝚢𝚘𝚞 𝚊𝚛𝚎 𝚗𝚘𝚠 𝚛𝚎𝚊𝚍𝚒𝚗𝚐)

BSD35k-CS_metadata.csv is the main metadata file, containing annotations and additional information for each sound. Each row corresponds to one sound and includes the following fields:

• sound_id: Freesound ID used as the unique identifier of the sound. The audio files found in the audiofolder are named using this ID, with a .wav extension for the audio format.
• class: Second-level class code of the sound.
• class_idx: Class index of 3 digits, with the first digit corresponding to the index of top level class, and the 2 last digits to the indix of second level class, both ordered according to the taxonomy.
• class_top: Corresponding top-level class code. It is derived from the full (second-level) class code by taking the part before the hyphen (-).
• confidence: This column is empty in this dataset. It is kept as part of the metadata file for data consistency with BSD10k.
• uploader: User who uploaded the sound in Freesound.
• license: Link to the license of the sound.
• title: Sound title provided by the uploader.
• tags: Tags associated with the sound provided by the uploader.
• description: Description of the sound provided by the uploader.

The mapping of class codes and class idx to their corresponding full class names can be found in BST_description.csv, which also includes a description and examples for each class. A diagram of the taxonomy (BST_diagram.png) is also included for a quick overview of the categories.

The features folder contains two subfolders with the aforementioned audio and text embeddings respectively.

Acknowledgments

This research is partially funded by the Generalitat de Catalunya (2023FI-100252, Joan Oró program), the IA y Música Cátedra (TSI-100929-2023-1, Cátedras ENIA 2022, SE Digitalización e IA, EU NGEU), and the IMPA project (PID2023-152250OB-I00, MCIU, AEI, co-funded by EU).