PDF Publication Title:
Text from PDF Page: 005
G Model POETIC 1239 No. of Pages 13 4 M. Airoldi et al. / Poetics xxx (2015) xxx–xxx title, description and tags (Cheng, Dale, & Liu, 2008: 236). Either way, as Google researchers Davidson and colleagues write, it seems fair to assume that the YouTube related videos algorithm is principally rooted upon “co-visitation counts” (i.e., co- views); additional data sources such as “time stamps of video watches” and other metadata are believed to be employed only in order to reduce “presentation biases” and “noise” in the resulting list of recommendations (Davidson et al., 2010: 294). This essentially means two things: a) YouTube videos can be seen as nodes in a network, with related videos inducing a directed graph in which an edge can be established between each pair of videos3 (Davidson et al., 2010: 295); and, b) the weight of the edges is determined by the users’ aggregated consumption practices on the platform (Bendersky et al., 2014). The action of co-viewing two media contents is shaped by technical elements, such as the presence of related videos themselves, as well as by the uploaders’ activity on the site. This may consist of curating YouTube channels, creating and naming playlists, discursively framing a song as ‘relaxing’ – thus enabling specific keyword-based exploration paths – or managing promotional activities by labels, concert venues, radio shows as well as ‘YouTube stars’ (see van Dijck, 2009). Up to now, despite the growing importance of Internet research (Coleman, 2010; Marres, 2012; Rogers, 2013), there have been very few studies in the social sciences that directly tackle online streaming services (e.g. Zhang et al., 2013). YouTube has been widely investigated through qualitative and quantitative approaches (see Giglietto, Rossi, & Bennato, 2012), but so far without considering the network of related videos. An interesting study of music categories on recommender websites Audioscrobbler.com and Musicmobs.com was conducted by Lambiotte and Ausloos (2005). The authors examined the large graph linking the users of the websites with online shared music libraries in order to “probe the reality of the usual music divisions, e.g. rock, alternative & punk, classical”, to propose a “quantitative way to define more refined musical subdivisions [ . . . ] that are not based upon usual standards but rather upon the intrinsic structure of the audience” (Lambiotte & Ausloos, 2005: 2). They detected clusters of artists sharing the same audience, thereby showing that many of these “islands” in the network corresponded to “standard, homogeneous style groupings” (Lambiotte & Ausloos, 2005: 6), such as country, dance, pop, swing, jazz, rock. Other clusters were “geographically localised”, and some revealed “unexpected collective listening habits, thereby uncovering trends in music” (Lambiotte & Ausloos, 2005). However, the authors assumed in principle the existence of a general correspondence between taste clusters and music categories – which has been increasingly questioned by omnivore-univore theorists in the sociology of taste (see Peterson, 2005). Furthermore, they focused on niche music publics – e.g. Radiohead was found to be the group with the largest audience (Lambiotte & Ausloos, 2005) – which makes their results not very generalizable to mainstream audiences. Our study of YouTube aims to overcome such limitations, since the associations among musical videos on YouTube are not determined by the existence of communities of users owning the same songs in their libraries, but by a multitude of micro- social practices – such as, building playlists and subscribing to channels – that generate aggregated co-viewing patterns. Also, YouTube’s worldwide ubiquity allows us to cover both mainstream as well as niche audiences. At the same time, we must consider that music on YouTube is rarely detached from video, and often the visual component is predominant over audio (Holt, 2011). Existing research has documented how music preferences are influenced by a variety of stimuli that include recommendation algorithms but largely remain a social construction. Tepper and Hargittai (2009) find in their study of music exploration habits by American college students that traditional social circles and mainstream media continue to be important means through which students learn about new music (2009:245). Arguably, it is not just technology that influences music reception, but it could also be the other way around. That is, most recommendation systems relate and distribute content by translating the behaviour of users into automated suggestions (Celma, 2010). Our research aims to expand the current knowledge on this process, and the implications this has for music classification. 3. DATA & METHODS This study is based on a mixed method approach that combines network analysis with content analysis taken together in a research design inspired by the ‘digital methods’ approach (Rogers, 2013). This latter approach proposes to take advantage of the way digital platforms produce and organise data in order to inform research around the analysis of large scale digital networks. This can be summarized by the motto ‘follow the medium’, which means to “natively” research the digital environment chosen, following the existing logic of action on the platform. In coherence with this methodological framework, we scraped, crawled and analysed a set of YouTube videos in order to reconstruct the network resulting from the aggregated consumption practices of millions of digital music consumers (Davidson et al., 2010). The data were retrieved using YouTube Data API v. 2.0,4 pulled on 26 February 2014. Data collection consisted of two phases. The first phase allowed us to obtain a generic list of videos related to music content, querying the API for the keyword ‘music’ and setting the language parameter to English. The extreme generality of the keyword allows for the inclusion of a 3 A network (or graph) is a collection of nodes (or vertices), which are the items being connected; edges (or links) are the connections between these items. 4 An API (Application Programming Interface) is a set of methods used to programmatically access a system. In this case, YouTube Data API allowed us to query YouTube’s servers for data related to music videos (a list of videos from a keyword and a list of related videos for each video). YouTube Data API version 2.0 has now been replaced by version 3.0. The documentation is still available here: https://developers.google.com/youtube/2.0/developers_guide_pro- tocol_api_query_parameters [Accessed: 10 January 2016]. Please cite this article in press as: M. Airoldi, et al., Follow the algorithm: An exploratory investigation of music on YouTube, Poetics (2016), http://dx.doi.org/10.1016/j.poetic.2016.05.001PDF Image | Follow the algorithm: An exploratory investigation of music on YouTube
PDF Search Title:
Follow the algorithm: An exploratory investigation of music on YouTubeOriginal File Name Searched:
AiroldiBeraldoGandini2016Preprint.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
| CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com | RSS | AMP |