The Eurovision Song Contest is about as old as Finnish television history, and throughout its history the competition has been widely discussed in media. Today, in addition to more traditional media, there are also some dedicated fans (and more professional youtubers/influencers, such as wiwibloggs) who follow national selection processes and actively produce content on Youtube and other social media. There’s also more and more Eurovision related data available (especially on participants, results and voting, but recently also on audio features of songs). Here I’ll discuss how musicologists, other researchers and ESC enthusiasts have approached the competition, and how data science and music information retrieval can be used in Eurovision related research.
Building national, cultural and musical identities
The European Song Contest can be seen as a unique TV spectacle, but also as an interesting cultural arena that can be examined from various viewpoints. Compared to the amount of media coverage, the amount of academic research on the contest has been quite scarce. The existing articles have represented various types of approaches from quantitative voting related studies (or studies aiming to examine reciprocal relations between European countries) to case studies on specific songs – often in relation to current political or cultural issues. My impression – after browsing through some articles I found – is that the Eurovision Song Contest has mainly interested researchers in the context of power, politics and national, cultural and musical identities. As an example, in Finland, Mari Pajala has extensively analysed how the contest has been used in building a sense of nationality and how television has been used as a media for shared cultural memory. However, at least until recently, there has not really been comprehensive research on musical aspects of songs – such as genres, song structures, or typical features like elevating modulations – although these have been widely discussed in the media. In research, these issues have mainly been discussed as part of specific case studies. It can of course be argued that there is no need for this, but I think it would be very interesting to compare some common comprehensions and attitudes toward Eurovision in relation to information that can be extracted through music information retrieval.
From voting analysis to musical feature analysis
Today, various kinds of Eurovision related data sources, databases and datasets are available (see the list in the end of this post for examples). Some have started as fan website -based sources, and some have been constructed for more research-oriented purposes. Traditionally, data science -based approaches have mainly been used to examine voting patterns, but as the methods have evolved, there are more and more possibilities for both, researchers and data-interested Eurovision fans, to examine issues that interest them. As an example, in Medium, there are descriptions of voting block -related interests, other type of data science -oriented interests, and some analysis that has been done based on musical features (Spotify’s API) with the aim of predicting the winner.
Today, there are also some research groups and ongoing academic research projects that have taken a step toward examining musical features of Eurovision songs. Interestingly, Joe Bennett (musicologist from the Berklee College of Music) and Simon Troup, were commissioned by Netflix to explore the characteristics of the songs that attract the most votes, and Bennett has already published some preliminary results on his website. Some interesting experiments have also been done in the University of Amsterdam (and Amsterdam Music Lab). A team of Dutch musicologists (including Janne Spijkervet) tried to compose the perfect Eurovision hit (embedded below) with the help of AI techniques, and Ashley Burgoyne and Henkjan Honing, have used music from the Eurovision Song Contest in their experiments in order to learn about what makes songs recognisable and how we remember music in the long term. All in all, it seems that there’s a growing interest toward these kinds of studies. It’ll be interesting to see whether the Netflix commissioned study was just a grace note related to a produced movie or will there be more broadcasting related interests toward the competition and/or the planned American spinoff? Are we about to have datasets that could be used to compare European and American musical preferences – or for some other purposes? This remains to be seen.
Above: Abbus · Can AI Kick It – ‘The Eurovision hit’, partly produced by Artificial Intelligence
Some academic publications/articles:
Bohlman, P.V. (2007). The politics of power, pleasure and prayer in the Eurovision Song Contest. Muzikologija : časopis Muzikološkog instituta Srpske akademije nauka i umetnosti, 2007(7), 39-67.
Pajala, M. (2006). Erot järjestykseen!: Eurovision laulukilpailu, kansallisuus ja televisiohistoria. Jyväskylän yliopisto. https://jyx.jyu.fi/handle/123456789/42891
Wolther, I. (2012). More than just music: The seven dimensions of the Eurovision Song Contest. Popular Music, 31(1), 165-171.
Some examples of Data Science/ MIR related approaches:
Instructional materials for music information retrieval (MIR) (Site maintained by Steve Tjoa)
Some sources for Eurovision related data:
https://github.com/Spijkervet/eurovision-dataset – freely-available dataset containing audio features, metadata, contest ranking and voting data
https://eschome.net – ESC database that can be explored through a user interface
https://github.com/adamprice97/EurovisionVotingBlocks – dataset for detecting Eurovision voting blocks
Photo: Opening act of ESC 2011 / Frédéric de Villamil from Paris, France, CC BY-SA 2.0, via Wikimedia Commons