Small version of the poster I put together for today’s pop-up research event at the Open University. Node size indicates total number of followers for SoundCloud users based in each city; arrows indicate where those followers come from (so far as we can tell); node colour indicates centrality to the network of these relationships (by eigenvector centrality). If you want more technical details, read last September’s post.
Back in June, we held our first public event, with live music performances, talks, and free food. The talks and performances were recorded and soon they will all be available online. We’re starting with Anna’s, Byron’s, and my introduction to the Valuing Electronic Music project as a whole. In this talk, we explain how and why we have been studying the value of electronic music, and reveal a little of what we’ve found out so far.
Also available from the Open University podcast site.
On 23 September 2014, I gave an invited presentation to a meeting of the Creative Data Club, organised by Sound and Music, the national agency for new music. Also speaking were Chris Unitt from One Further, presenting a study on how arts organisations are using Facebook, Jay Short from inition, showcasing 3D-printed visualisations of social media data, Dan Simpson, talking about his crowdsourcing of poetic composition, and shardcore, telling the hilarious and poignant tale of Alex the twitterbot. Here are the slides, plus a few audience responses and livetweets at the end. The text is based on the same handwritten notes that I extemporised from on the day.
(Photograph by Sound and Music)
Sometimes we just want to get a simple overview of the types of things people are saying. In the case of our SoundCloud analysis, we want to know what people are saying about each other’s tracks.
We’ve made use of http://www.wordle.net/ word clouds to get an overview: what words are people typically using in comments on SoundCloud? Are they positive? descriptive? critical? irrelevant?
Wanting to find out what was typical SoundCloud behaviour – as opposed to what our case study users were doing – we took a random sample of 150000 SoundCloud accounts earlier this year and downloaded their profile data, plus the profile data of everyone they were following (plus some other stuff, but that’s for another time). One of the things we did with this data was to construct a social network graph showing ‘follow’ relationships at city level: every time our computer program found that a sampled user self-identified with city A followed a user self-identified with city B, it created an ‘arc’ (represented with an arrow) from city A to city B. We then combined all the arcs so that instead of, say, 2000 arcs from city A to city B, there would now be a single arc with a ‘weight’ of 2000. We then imported this data into Gephi, sized the nodes representing cities to reflect the total weight of all the incoming arcs, positioned them with the Force Atlas algorithm, and used the Louvain community detection method to identify ‘clusters’, where a cluster is a group of nodes that are better connected to each other than they are to nodes from outside the group. And here’s the result, with five colours to represent the five clusters.
In my previous post on this topic, I introduced a problem – how to understand the work that explicit genre categorisations are made to do by people uploading tracks to the SoundCloud audio-sharing website – and a potential solution – identifying the three categories most frequently used by each individual in a sample and studying regularities in the ways in which pairs of categories tend to pop up within the same group of three. I also presented some partial and preliminary findings in the form of a matrix comparing co-occurrences of the five genre categories most frequently used by people within an initial sample. And I either glossed over or left unmentioned a slew of problems, some of which we’ve been more successful in addressing than others at present (because these are only blog posts, and we haven’t finished the research yet). The biggest problem is the sample itself: the analysis was done on the basis of a snowball sample, when a random sample would be more appropriate. Hence the provisionality of all this. The analysis will be redone soon on the basis of a sample that will enable us to make more robust claims, but in the meantime I wanted to share our thought processes and working methods with the world because – quite apart from anything else – I’m excited about the patterns that are emerging.
We have been exploring how visualisations can illustrate over time how users comment on tracks in SoundCloud. Commenting has been highlighted in our qualitative research as a way of building relationships and showing appreciation of other musicians’ work. In fact, initial inspection of a sample set of comments is showing that most comments in our samples tend to be positive or constructive, rather than overly critical or negative.
We have created two visualisations:
It’s a scary thought, but we’re halfway through the funded stage of this project. The timescale is tightly compressed and it’s been a bit manic at times (especially right now, with a workshop next week and a public event in less than a month).
But we’ve already learnt so much. So I’d like to reflect briefly on an observation Byron made last month, reflecting on the interviews he’d been carrying out with electronic music producers: ‘when I ask questions about valuing and appreciation, people answer about relationships.’ Outside the spheres of commercial music (where value is expressed in economic terms) and art music (where it is expressed in terms of grants, academic appointments, etc), human relationships are the beginning and the end of musical value. So perhaps it’s true that when musicians build relationships through music, they are producing not the opportunity to produce value (the aim of ‘business networking‘), but value itself.
As >our qualitative research starts, so does our quantitative research. We have been collecting data from SoundCloud’s API (Application Programming Interface) which is the gateway to access SoundCloud’s data. We have also been updating the IF analysis code written by Daniel Allington for analysing network actions in Interactive Fiction communities, to make it useful for analysing what happens between users on SoundCloud. SoundCloud has made available an SDK (Software Development Kit) for Python and other programming languages, which is a set of functions and programs that we can use in our code to do things with SoundCloud data.