Alo Allik, Centre for Digital Music, Queen Mary University of London
MusicLynx is a web application for music discovery that enables users to explore an artist similarity graph constructed by linking together various open public data sources. It provides a multifaceted browsing platform that strives for an alternative, graph-based representation of artist connections to grid-like conventions of traditional recommendation systems. Bipartite graph filtering of the Linked Data cloud, content-based music information retrieval, machine learning on crowd-sourced information and Semantic Web technologies are combined to analyze existing and create new categories of music artists through which they are connected.
There are a number of openly accessible music-related knowledge bases and datasets, that crowd-source different types of information and make it available to the community. The MusicLynx API uses and makes connections between MusicBrainz, Dbpedia, AcousticBrainz, Million Song Dataset, Last.fm, BBC Music, Wikidata and others. The similarity graph displayed to users is built utilizing graph filtering methods, including Collaborative Filtering, Maximum Degree Weighted, and hybrid Heat/Probability Spreading that enables balancing between accuracy and diversity of similarity.
The MusicLynx application consists of two components: the Angular2 front-end that serves content to users and the MusicLynx API implemented in ExpressJS which consists of modules that query the different services for data and process it. MusicLynx is available online: http://musiclynx.github.io/
MusicLynx API is accessible as a stand-alone public interface: http://musiclynx-api.herokuapp.com/
Both components are open-source and published under GNU General Public License v3.0.