WikiMuTe: A web-sourced dataset of semantic descriptions for music audio

This upload contains the supplementary material for our paper presented at the MMM2024 conference.

Dataset

The dataset contains rich text descriptions for music audio files collected from Wikipedia articles.

The audio files are freely accessible and available for download through the URLs provided in the dataset.

Example

A few hand-picked, simplified examples of the dataset. 

Dataset structure

We provide three variants of the dataset in the data folder.

All are described in the paper.

1. all.csv contains all the data we collected, without any filtering.
2. filtered_sf.csv contains the data obtained using the self-filtering method.
3. filtered_mc.csv contains the data obtained using the MusicCaps dataset method.

File structure

Each CSV file contains the following columns:

• file: the name of the audio file
• pageid: the ID of the Wikipedia article where the text was collected from
• aspects: the short-form (tag) description texts collected from the Wikipedia articles
• sentences: the long-form (caption) description texts collected from the Wikipedia articles
• audio_url: the URL of the audio file
• url: the URL of the Wikipedia article where the text was collected from

Citation

@inproceedings{wikimute,
    title = {WikiMuTe: {A} Web-Sourced Dataset of Semantic Descriptions for Music Audio},
    author = {Weck, Benno and Kirchhoff, Holger and Grosche, Peter and Serra, Xavier},
    booktitle = "MultiMedia Modeling",
    year = "2024",
    publisher = "Springer Nature Switzerland",
    address = "Cham",
    pages = "42--56",
    doi = {10.1007/978-3-031-56435-2_4},
    url = {https://doi.org/10.1007/978-3-031-56435-2_4},
}

License

The data is available under the Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license.

Each entry in the dataset contains a URL linking to the article, where the text data was collected from.