MusicWeb: music discovery with open linked semantic metadata

Abstract

This demo presents MusicWeb, a novel platform for linking music artists within a web-based application for discovering associations between them. MusicWeb provides a browsing experience using connections that are either extra-musical or tangential to music, such as the artists' political a liation or social in uence, or intra-musical, such as the artists' main instrument or most favoured musical key. The platform integrates open linked semantic metadata from various Semantic Web, music recommendation and social media data sources. The connections are further supplemented by thematic analysis of journal articles, blog posts and content-based similarity measures focussing on high level musical categories.

Keywords

• Semantic Web
• Linked Open Data
• music metadata
• semantic audio analysis
• music information retrieval

Introduction

MusicWeb is a music discovery platform which o ers users the possibility of exploring editorial, cultural and musical links between artists. It gathers, extracts and manages metadata from many di erent sources. The connections between artists are based on YAGO categories such as style, geographical location, instrumentation, record label, but also more obscure links, for instance, artists who have received the same award, have shared the same fate, or belonged to the same organisation or religion. The connections are further enhanced by thematic analysis of journal articles and blog posts, content-based music information retrieval similarity metrics and proximity measures in a 2-dimensional mood space. Information about artists is collated and processed from several different web knowledge content resources and presented for the user to navigate in a faceted manner [ 5 ]. MusicWeb as a discovery platform is subtly di erent in function from recommender systems like dbrec [ 7 ], which suggests music obtained from DBpedia by computing a measure of semantic distance as the number of indirect and distinct links between resources in a graph. This demo intends to show the methods of linking artists employed in the system and how these could help overcome issues such as infrequent access of lesser known artists in large music catalogues (the \long tail" problem) or the di culty of recommending artists without user ratings in systems that employ collaborative ltering (\cold start" problem) [ 3 ]. 1 http://musicbrainz.org 2 http://dbpedia.org 3 http://sameas.org 4 http://last.fm for linking artists. the Music Ontology [ 8 ] provides main concepts and properties for describing musical entities, including artists, on the Semantic Web.

AcousticBrainz5 service, which gathers crowd-sourced acoustic information about music, facilitates content-based similarity calculation. 3

Artist similarity

There are many ways in which artists can be considered related. MusicWeb uses Semantic Web technologies and linked data to facilitate faceted searching and displaying of information [ 6 ]. This is done by modeling artist similarities in four di erent domains: socio-cultural, research and journalistic literature, crowdsourced tag statistics and content-based information retrieval.

Socio-cultural connections between artists in MusicWeb are primarily derived from YAGO categories that are incorporated into entities in DBpedia. Many categories, in particular those that can be considered extra-musical or tangential to music, stem from the particular methodology used to derive YAGO categories from Wikipedia [ 4 ].

Literature-based linking is achieved by data-mining research articles and online publications using natural language processing. MusicWeb uses Mendeley6 and Elsevier7 databases for accessing research articles that are curated and categorised by keywords, authors and disciplines. Online newspapers, music magazines and blogs, on the other hand, constitute non-curated data. Relevant information in this case must be extracted from the body of the text by Web-crawling based on keywords or tags. The Alchemy API8 is then used for named entity recognition and keyword extraction.

Crowd-sourced tags enable modelling similarity based on projected mood. This method involves using the Semantic Web version of ILM10K music mood dataset that consists of over 4000 unique artists [ 1 ]. The dataset is based on crowd-sourced mood tag statistics from Last.fm users, which have been transformed to 2-dimensional coordinates re ecting energy and pleasantness. The similarity between artists is measured by rst obtaining the average location of each artist based on their track coordinates. The average locations then enable computing distances between artists and using these as the similarity metric. Content-based linking involves methodology of Music Information Retrieval (MIR) [ 2 ] which facilitate applications that rely on perceptual, statistical, semantic or musical features derived from audio using digital signal processing and machine learning methods. These features may include statistical aggregates computed from time-frequency representations extracted over short time windows. Higher-level musical features include keys, chords, tempo, rhythm, as well as semantic features like genre or mood, with speci c algorithms to extract this information from audio. To exploit di erent types of similarity, we 5 https://acousticbrainz.org/ 6 http://dev.mendeley.com/ 7 http://dev.elsevier.com/ 8 http://www.alchemyapi.com model each artist using three main categories of audio descriptors: rhythmic, harmonic and timbral. The features are obtained from the AcousticBrainz Web service which provides descriptors in each category of interest. For each artist in our database, we retrieve features for a large collection of their tracks in the above categories, including beats-per-minute and onset rate (rhythmic), chord histograms (harmonic) and MFCC (timbral) features. 4

Conclusion

MusicWeb is an emerging application to explore the possibilities of linked databased music discovery. It facilitates users to engage in interesting discovery paths through the space of music artists. The aim is to gather in one application various di erent approaches to music discovery and how they can bene t from linked music metadata. The next steps are directed toward evaluating its potential acceptance by end users, in particular, exploring which linking methods listeners nd most appealing or interesting, and which they would use more often. MusicWeb is accessible online: http://musicweb.eecs.qmul.ac.uk/

References

1. Mathieu Barthet , Gyorgy Fazekas, Alo Allik, and Mark B. Sandler . Moodplay: an interactive mood-based musical experience . In George Kalliris and Charalampos Dimoulas , editors, Proceedings of the Audio Mostly 2015 on Interaction With Sound , AM '15 , Thessaloniki , Greece, October 7- 9 , 2015 , pages 3:1{3:8 . ACM, 2015 .

2. Michael A. Casey , Remco Veltkamp, Masataka Goto, Marc Leman, Christophe Rhodes, and Malcolm Slaney . Content-based music information retrieval: Current directions and future challenges . volume 96 . IEEE Proceedings, April 2008 .

3. O. Celma . Music Recommendation and Discovery:The Long Tail, Long Fail, and Long Play in the Digital Music Space . Springer Verlag, 2010 .

4. M.S. Fabian , K. Gjergji , and W. Gerhard . YAGO: A core of semantic knowledge unifying wordnet and wikipedia . In 16th International World Wide Web Conference, WWW , pages 697 { 706 , 2007 .

5. Gary Marchionini . Exploratory search: from nding to understanding . Communications of the ACM , 49 ( 9 ), 2006 .

6. E. Oren , R. Delbru , and S. Decker . Extending faceted navigation for RDF data . In ISWC , pages 559 { 572 , 2006 .

7. Alexandre Passant . dbrec - music recommendations using DBpedia . In Peter F. Patel-Schneider , Yue Pan, Pascal Hitzler, Peter Mika , Lei Zhang, Je Z. Pan , Ian Horrocks, and Birte Glimm, editors, The Semantic Web - ISWC 2010 - 9th International Semantic Web Conference, ISWC 2010 , Shanghai, China, November 7- 11 , 2010 , Revised Selected Papers , Part II , volume 6497 of Lecture Notes in Computer Science, pages 209 { 224 . Springer, 2010 .

8. Yves Raimond , Samer A Abdallah, Mark B Sandler, and Frederick Giasson . The music ontology . In ISMIR , pages 417 { 422 . Citeseer , 2007 .