IDSC Big Data Analytics & Data Mining Postdoctoral Researcher Mohamed Sordo has launched Musikipedia.org, a powerful web application for analyzing and streaming music. Musikipedia stands at the forefront of music analysis, illustrating the potential that Big Data could have on our everyday lives.
Utilizing a technology called NLP, or natural language processing, to interpret and process meaningful data about music, this new website masterfully navigates Wikipedia’s immense knowledge base and extracts machine-readable data. It then analyzes and draws data relations between different musical works, aggregating attributes of music such as the artist’s birthplace, country of citizenship, hometown, and occupation. It also detects relationships between other artists and reveals associated record labels, artists, and bands. In addition, Musikipedia analyzes audio content to determine, for example, musical genre, and to unearth corollary lesser-known artists. This data-driven approach to music opens up an exciting future of expanded discovery for the casual listener and music enthusiasts.
Sordo created Musikipedia using his insight into music, computer science, and data mining. He saw that while popular music streaming services like Pandora and Spotify provided access to a vast array of music, their interactivity with users was limited. Sordo sought to harness the potential of the abundance of music-related data available on the Internet to reach beyond the face value of music and dig deeper, focusing on the context of music.
As Sordo explains, current music technology encourages users to engage in ‘passive listening,’ which limits their participation in the music discovery process. Until Musikipedia, those eager to discover new artists and new music were limited to very basic metadata such as track name, artist, and album. Now you can, for example, listen to other bands from the same place, same record label, or same producer. Musikipedia allows users to take part in a much more engaging and interactive way to listen to music—a kind of ‘stumble-upon’ approach that facilitates discovery. Musikipedia is a powerful new tool, providing a more enjoyable and educational experience for listeners, and with the potential to contribute to the creative process for artists.
-by Nicolás Aguirre
Musikipedia is a music service that allows you to browse and discover music using information extracted from Knowledge Bases.
More specifically, it uses Wikidata, DBPedia, Wikipedia and MusicBrainz as its backbone. Currently, it extracts information for 4 types of entities: songs, works, albums and artists; this will be extended in the near future to include other music-related entities (such as instruments, genres, moods, etc.). Moreover, songs (and works) are matched with a youtube video, so that the user can listen/watch the song while reading/browsing content.
One of the distinguishing characteristics of musikipedia (when compared to other music services such as Last.fm) is that it relies on the categories and attributes defined in the aforementioned Knowledge Bases to create a list of related entities. For example, songs The Streets by WC and Clap Back by Ja Rule are related (according to musikipedia) because they share producer and writer (Scott Storch), music genre (Hip Hop music) and record label (Def Jam Recordings).
As for the search, it is still not in its full capabilities, but for example one can query “produced by Scott Storch” and get an attribute page with song recordings produced by Scott Storch. Obviously, searching for a specific artist or song works as well.
IDSC “Where Are They Now?”
After leaving IDSC in 2016, Mohamed moved to the California Bay Area to join Pandora.com as a Data Scientist, where he applies Machine Learning, Information Retrieval, and Natural Language Processing techniques to recommend music at scale to millions of Pandora listeners. Pandora Internet Radio is a music streaming and automated music recommendation service powered by the Music Genome Project, operated by Pandora Media, and available only in the U.S.