Learned Genre Ontologies

Based on the method proposed in [1], we have learned two genre ontologies (OEng and OIntl) from crowdsourced genre labels submitted to the central beaTunes database. Both OEng and OIntl are described in great detail in [1]. Though it has nothing to do with the Million Song Dataset (MSD), the method employed is a refinement of the approach taken for the tagtraum genre annotations for the Million Song Dataset.

Due to the nature of the creation process, the ontologies contain errors like noise/garbage (e.g. !scott), multiple entries for the same thing (translations), and genres labeled as Music Genre, that clearly are not music genres (e.g. Book).

You can access the generated ontologies via the links below. All data is presented in human readable HTML as well as in RDF-XML, RDF-JSON, JSON-LD, and Turtle.

Name Namespace Description Link
OEng https://www.tagtraum.com/genres/eng/2016-02-02/ Generated from the top 1,000 genres of English speaking users. Go
OIntl https://www.tagtraum.com/genres/intl/2016-02-02/ Generated from the top 400 genres of English, Dutch, German, French, and Spanish users. Go

Additionally to the originally published ontologies (namespace 2016-02-02), we have released a second set of ontologies. These are based on a newer and larger snapshot of the beaTunes database. Also, instead of using the top 1,000 genres as described in [1], we used the top 2,000 genres for the ontologies based on language specific users and the top 800 genres/language for the international ontology.

Name Namespace Description Link
OEng https://www.tagtraum.com/genres/eng/2016-11-11/ Generated from the top 2,000 genres of English speaking users. Go
ODeu https://www.tagtraum.com/genres/deu/2016-11-11/ Generated from the top 2,000 genres of German speaking users. Go
OFra https://www.tagtraum.com/genres/fra/2016-11-11/ Generated from the top 2,000 genres of French speaking users. Go
OEsp https://www.tagtraum.com/genres/esp/2016-11-11/ Generated from the top 2,000 genres of Spanish speaking users. Go
ONld https://www.tagtraum.com/genres/nld/2016-11-11/ Generated from the top 2,000 genres of Dutch speaking users. Go
OIntl https://www.tagtraum.com/genres/intl/2016-11-11/ Generated from the top 800 genres of English, Dutch, German, French, and Spanish users. Go

Content Negotiation

The server supports a simple form of content negotiation via 303 redirects, based on HTTP Accept headers.

For example, The following line will download the RDF-XML representation of the index resource, which contains all statements:

curl -L -H "Accept: application/rdf+xml" \
  https://www.tagtraum.com/genres/eng/2016-02-02/index

To fetch statements about individual genres (e.g. Rock), use something like this:

curl -L -H "Accept: application/rdf+xml" \
  https://www.tagtraum.com/genres/eng/2016-02-02/rock

Alternatively, you can access the representation you want by appending the desired extension suffix. E.g. .jsonld for JSON-LD.

FAQ

What is the licensing?

Creative Commons License This work is licensed under a Creative Commons Attribution 4.0 International License.

How to cite the ontology?

Please cite this publication (pdf):

[1] Hendrik Schreiber. Genre Ontology Learning: Comparing curated with crowd-sourced ontologies. In Proceedings of the 17th International Society for Music Information Retrieval Conference (ISMIR), New York City, USA, Aug. 2016.

Other research.