Javier Trujillo Garcia, Jeroen de Vos, Oscar Coromina, Reinier Tromp, Simeona Petkova
This project pertains to the question of how technologies (in our case the platform SoundCloud) mediate memories since this is a field (as an intersection between media and memory studies) that needs to be further explored. Emotions are vital to the processes of remembering and there is a strong interrelation between music and emotions, therefore researchers have argued that music is vital to memory as well. Bijsterveld and Van Dijck (2009) acknowledge that people either consciously or involuntarily can recall past events and emotions through sound and music: ...people make use of audio technologies to elicit, reconstruct, celebrate, and manage their memories, or even a past in which they did not participate. (Bijsterveld and Van Dijck: 11). Music relates to many emotional events that individuals can experience (Jäncke, 2008). When autobiographical memories are recalled, the person may feel similar emotions to those felt at the actual event. In addition, most individuals have extensive and lifelong experience hearing music as part of both everyday life and special events (Scherer and Zentner, 2001). Thus, music can evoke emotions due to either association with personal events or through the intrinsic emotional nature of the music itself (Krumhansl, 2002).
Moreover, the understanding of how music shapes our perception of past and present has been complicated by the evolution of sound recording and the emergence of digital techniques has greatly accelerated the use of musical archives as symbolic and material resources within the composition of music, as well as a seemingly endless dissemination of music files.
In recent years, SoundCloud has established itself as a platform for such music related practices. Our interest in this platform has been further fueled by its presence as a node in the (interactor) network of commemoration of the 10th anniversary of John Peels death. John Peel is considered to be one of the most influential radio DJs in UK and around the world, having forty years of broadcasting practice for BBC, subjecting his radio career to introducing and airing new, independent, different and progressive music. His sudden death in 2004 has sparked a public debate how he should/should not be remembered: do the books, autobiographies, collections of his radio Sessions fulfill their purpose as memory tools or is naming a stage at Glastonbury festival or train after him too excessive? Parallel to the official rites of remembering, an alternative has been established in the form of an online movement of keeping Peels music legacy alive.
Therefore, the aim of this project is to investigate the specificity of Soundcloud as a music sharing platform (in relation to memory dynamic) and to map the network of Peels commemorative practices enabled by SoundCloud.
Figure 1 John Peels Interactor Map around 10th anniversary of his death
Previous platform specific researches:
There has been little research done on the SoundCloud platform and the most informative studies have been done in Sweden. This is by no means a coincidence, for SoundCloud was founded by the two Swedish entrepreneurs Alexander Ljung and Eric Wahlforss. Unfortunately both bodies of literature are only published in the original language, Swedish, which made it a bit harder to look into the actual results and methods used. Nevertheless a short description will give an overview of the two different studies.
Patrick Stanelius, Master Student in Journalism, provides a more theoretical (and self-ethnographic evaluation of SoundCloud. In the paper I ljudmolnet: Om identitet och ungdomskultur på Soundcloud [Sound cloud: Identity and Youth Culture on Soundcloud] published by the University of Stockholm, he shows how social interaction on SoundCloud has affected users' perception of the music and in a depth plan, including claims to SoundCloud to resemble with the idea of a community (Stanelius 2015, 4). In turn the study lead to an initial ethnographic enquiry in different usages of the of the platform. The paper examines what binds users together in the act of using SoundCloud exclusively to find new music and share their own (uploaded) music (Stanelius 2015, parapharase mine).
The second research lacks the theoretical introduction, but instead goes into ethnographic detail on finding and filtering mechanism. In the paper Underground in the Cloud: En kvalitativ studie om den digitala musikplattformen Soundcloud [Underground in the Cloud: A qualitative study of the digital music platform Soundcloud] Adam Kuylenstierna builds upon a rigid set of emperical data. The author indicates that SoundCloud can be considered a step forward in the process of making the act of sharing music more social (Kuylenstierna 2011, 2; translation mine). However, and this quite relevant to our project, the sharing development of the sharing process might actually be slowed down for the ties between users are in many cases too weak. Kuylenstierna observes that the weak ties might actually be evoked by a filtering and selection mechanism that is too powerful, thereby stimulating sever filtering over community interaction (idem).
How is the dynamic of remembering / forgetting enabled through Soundcloud, concerning:
The frontend of the website
The back-end of the website / through the API
How can we map John Peel's memory network on Soundcloud?
To answer the question on remembering / forgetting John Peel we looked at the tracks that were related to John Peel. For getting the data from soundcloud we used the soundcloud API / Developers toolkit that is available once registered. The API from soundcloud provided us with unstructured data. Therefore we used google and Google Scraper to demarcate the tracks within Soundcloud related to John Peel. This provided us with a list of 258 unique tracks.
Initially we looked at different playlists in the Soundcloud API, however due to problems with retrieving proper data of playlist basis, we decided to step one level down and focus on individual tracks. We took a comparative approach investigating in the difference between tracks in the Soundcloud API and the database used by Google. We did this in order to get insights into the potential (temporal) differences between the front-end cached by Google and the back-end provided through the API. Some of the track indexed by Google were actually not available through either front- or back-end. Therefore we concluded that Google probably stores the data of soundcloud longer than the API. We wanted to know if there is a discrepancy between google's actual data, the data from the Soundcloud API and the website of Soundcloud to see if songs that has been deleted really have been deleted. We used the censorship tool for this investigation.
This tracklist was used to query the Soundcloud API for the list of tracks. This was outputted in a csv.file and analysed. The dataset was generated by tracks. Because Soundclouds content is user-generated, users itself are responsible for the tags, comments and the genre of the tracks.To look at the categorization of the genres and the tracktypes, and look at with what kind of music John Peel is remembered, we used the variables genre and tracktype. The results were visualized in in a pie chart.
Additionally, to find out the existence of a community around the person of John Peel on Soundcloud we looked at the connection between users related to John Peel. Every track has a unique id and a unique user with a user id. The variables we used to look at the relation between users and tracks were comments and tags. For this network analysis we used Gephi. The first relation we analysed was between tracks and comments. To see if people were relating to John Peel on the same subjects we made a second network analysis on tracks and tags, also using Gephi.
After accessing soundcloud via the API we found that a significant portion of the fields related to the track object were actually empty. Those fields were mostly the ones in which users are supposed to introduce information to better describe the uploaded content. Figure 1 shows as a matrix all the information we gathered via the API in which the fields colored in red corresponds to empty metadata fields. This emptiness paired with messiness regarding the format of the information contained in the field, by which the expression dirty database comes really handy to characterize soundclouds API.
Figure 2 the red fields represent a void in data
Another problem relating to the API had to do with the deepness of the data retrieved when asking for a single tracklist. We did not only get the general information of the tracklist but also all the information of every track that composed it. Additionally, since tracks are created by users, we would get all the information related to each author. As a result, we ended up with a large amount of data which was complicated to convert into a proper format ,such as .csv or .tsv, in order to perform statistical analyses and visualizations. This same problem also arose when asking for information about the comments of a particular track, in which the resulting .csv conversion resulted to have a huge and undefined number of comments displayed horizontally.
Additionally, we found a significant inconsistency in the way users tag and define the genre of every track. In the case of the tags, which are supposed to be a tool to organize and access similar contents, almost every track analyzed has its own set of unique tags. We can see that on the track-tag network illustrated in figure 2 in which tracks are colored in blue and tags in red. There are only a few tags shared between tracks being John Peel the most prominent. Interestingly, the tag john is another prominent tag. A fact that arises questions about the efficacy of soundcloud tags in terms of organizing similar music and also regarding the retrievability of the content.
A similar pattern appears when we analyze the genres of the 250 tracks in order to find for which musical genre John Peel is most remembered. As the pie chart in figure 3 shows, most of the tracks are related to a unique genre and attuned with the inconsistency and lack information we previously described, the most recurrent categorization is the empty field.
The difficulty of finding relations between soundcloud objects appear again when we try to find the community involved in the remembering of John Peel. Figure 4 reflects the network of commenters (red nodes) and tracks (green nodes) or, to be more precise, the absence of this network. Similarly as in the tag-track network there are only a few users that comment in at least two different tracks. This fact strongly suggests the absence of such community and, therefore, makes really difficult to consider Soundcloud as a device that plays a role in keeping alive the memory of John Peel legacy.
Finally, using Google as an entry point to obtain our list of tracks and checking the urls for 404 give us a list of 3 tracks that werent accessible anymore in soundcloud. Cross Checking those urls via the API we found that they were actually deleted -inaccessible-. Therefore, at the moment one track is deleted ceases to be reachable both in the frontend and the backend.
During the presentation, we made the quite bold statement that perhaps, Soundcloud could be considered a 'post-network' network. It builds upon our findings of networks of users, tags and tracks being quite isolated from each other. One critical limitation of our findings that needs mentioning here is that we took the 'track' as a starting unit of analysis. While trying to outline the networks of users commenting on multiple tracks and the network of John Peel related tracks sharing a consistent tag, we ran mainly into isolated nodes.
This brings up the question to what extend SoundCloud is actually a networked platform? The isolated nature of the back-end of the data suggests SoundCloud to be rather a tool than a networked platform; it provides a mediator between groups of people that can hardly be called a group. The bold term post-network network would imply a platform that mediates and networks itself, but which by design does not network different sites of data. Note that coining this term is done by means of facilitating a discussion rather than making a rigid claim on SoundCloud.
The second broader discussion points is a direct result of the data we got from the API. Looking a the output file of the API the data on the backend of SC seems to be quite unstructured both of the level of organization, filtering and structuring internal data, and on the level of the API interface giving messy output. Rather than a lack of actual communities present on soundcloud, the lack of findings might actually be explained though the fact that the data gathered by SoundCloud might actually be too chaotic to be useful for either Soundclouds internal, or researchers external purposes.
At last, maybe there might not be a John Peel memory community on Soundcloud. This does not mean that there are no act of remembrance to be found on SC, rather the act of commemorating the legendary radio maker in our research seems to be limited to individual attributions to the John Peel repertoire in forms of tagging, gerne distribution and comments. These performances just do not consolidate into a networked form.
Bijsterveld, K. and Van Dijck, J. eds.(2009). Sound Souvenirs: Audio Technologies, Memory and Cultural Practices.Amsterdam: Amsterdam University Press.
Jäncke, L. (2008). Music, Memory and Emotion. Journal of Biology 2008, 7:21.
Kuylenstierna, A. (2011). Underground in the Cloud: En kvalitativ studie om den digitala musikplattformen Soundcloud [Underground in the Cloud: A qualitative study of the digital music platform Soundcloud]. (Student paper). Stockholms universitet.
Stanelius, P. (2015). I ljudmolnet: Om identitet och ungdomskultur på Soundcloud. [Sound cloud: Identity and Youth Culture on Soundcloud] (Student paper). Stockholms universitet.Zentner et al. (2001). Emotions Evoked by the Sound of Music: Characterization, Classification, and Measurement. Emotion. 7: 4, 494 521