URL: Last.fm
Category: Music Distribution/Music Taxonomy/Sociable Media
Location: United Kingdom
Brief Description:
Last.fm is an music streaming engine powered by a dynamic relational database that develops according to the listeners music habits.
Why is this of Interest:
Last.fm is a very interesting example of a powerful and recursively populating (and generative) database system. last.fm is a commercial consumer engine that if viewed through the lens collaborative media provides many essential clues to building a powerful collaborative interface and ecosystem. While predating the so-called Web2.0 phenomena it exemplifies the power of the productive exchange of data for functionality. It provides an interesting take on the building a dynamic relational system based a very focussed, immediate and ‘affect’ driven taxonomy (listened,loved,banned) that structures and opens onto a very dynamic media ecology building relationships between users to auto-populate a dynamic and generative topology of music consumption. The ease of populating a relational database with a client based automated submission system is also of real interest.
Development History:
There are two streams to the Last.fm development history. These streams officially merged in 2005 although there integration began around 2003 according to the Wikipedia entry (http://en.wikipedia.org/wiki/last.fm). The two streams include the development of Audio Scrobbler on one hand and the Last.fm streaming engine on the other. Audio Scrobbler was an application that built a database of a users music consumption habits developed by Computer Science student Richard Jones (University of Southampton). These consumption habits were then folded into a recommendations engine that made connections between artists based on the profiles of the users that ‘scrobbled’ their listening history. An open API encouraged the development of plug-ins that allowed users to ‘scrobble’ listening data automatically via the music software of there choice. The integration of the iPod and the iTunes environment also provided the incipiency for the playing history of the iPod to be ‘scrobbled’ automatically. The ease of uploading data, the ubiquity of compatible music software and hardware, meant that Audio Scrobbler was able to develop a massive relational database and a huge and very active community of ‘scrobblers’. Last.fm is now the ‘face’ of audio scrobbler which has become completely integrated into the last.fm streaming engine.
Last.fm was founded in 2002 by Felix Miller, Martin Stiksel, Michael Breidenbrueker, Thomas Willomitzer. Last.fm is modeled was an internet radio station and recommendations engine that developed user profiles similar to those developed by Audio Scrobbler. The Last.fm engine was however able to actively build a playlist based on those profiles and stream that playlist to its users. The dynamically generated playlist that was essentially beyond the ability of the user to select what they were listening to meant that this model survived the wrath of the RIAA and other copyright stakeholders who objected to user prescribed streams. This will potentially change with a massive disruption to the fee-structure for copyright payments recently announce. Last.fm is a subscription based service that asks for a donation in order to unlock more ‘prescriptive’ play-listing. This revenue pays for radio licensing. While on the original Last.fm developed profiles based on a simple ‘affective’ taxonomy based on categories chosen as the user listened (love, skip, ban) subsequent development of the site/engine has included many elements borrowed from the social networking phenomenon (facebook, myspace, and earlier models). Listening ‘groups’, ‘friends’,'tags’ all began to ‘dilute’ the underlying ‘affective’ taxonomy (see my analysis). The top track this week on Last.fm is Mika’s Grace Kelly it was ‘scrobbled’ 52,147 times this week by 15,685 listeners. The same track has been scrobbled 445,920 times in total.
Operation/Analysis:
Last.fm is based on the fairly simple back end of a relational database in which user’s developing profiles provides links between tracks and artists dynamically. A layer of social networking has developed over the top of this engine and sometimes threatens to obfuscate its original power. There are many lessons to learn in structuring an interactive, transductive, database from last.fm’s development.
The original power in last.fm engine was its ability to move beyond the performance aspect that usually operates in any online identity engine. By ‘scrobbling’ data automatically based on the users actual listening history the engine avoided the usual quirks of a more ‘reflexive’ self-profiling. On the early last.fm/audio scrobbler integrations (before the development of social bookmarking features) you ‘were what you listened to. The addition of the simple last.fm schema of skipped, loved and banned options meant that the data was being ‘scrobbled’ according to an affective reaction to a track, the simplicity of the schema reduced the tendency to ‘intellectualize’ this reaction. This meant that Last.fm escaped the vagaries of a ‘tag’ based network. In a tag based network we as users are moved to log a tag to a object but the tag itself is much more dependent context etc. On last.fm tags were effectively delimited to those three ‘affective’ categories and the relational power of last was built on the connectivity that such a delimited schema provided between profiles and consequently between artists, between individual tracks etc.
Although streaming is obviously delimited to the tracks available to last.fm the DB accepts entries from any artist or recording at all. This makes the engine particularly capable of handling and encouraging cultural diversity. While many engines would simple not be able to deal with obscure data objects Last.fm, simply integrates that data as a connective tissue in the recommendation network. I can now be connected to the ears (through the affective ‘interface’) of the small group of users that listen to Tim Hecker or Ktu despite the fact that prior to the ‘scrobbling’ of that data last.fm had no information on these artists. Because of the simple interfacing this recommendation engine ‘simply works’ there are no ‘false’ submissions.Last.fm works on simple and largely autonomic recursion between listening habits and recommendations. Simply listening generates a profile which folds into the development of a play-list.
The Last.fm engine provides for the ongoing emergence of a dynamic topology of musical listening habits. This stands in direct opposition to the other engine that Last.fm is often grouped with the other important example of dynamic streaming ‘Pandora’. Pandora employs listening experts and an automated algorithm (largely musicians) to break down the characteristics of an audio/music file based on tempo, instrumentation, style and so on (see Pandora entry). Pandora presents another model of database entirely. The opposition is one of typology to topology. My preference is for the latter but it should be noted that the latter is only effective if the ‘topos’ is effectively ‘wrangled’ – if we look at a system like del.icio.us we find a topology literally gone mad in that the difference between tags and ease of tag production makes logical connections between users based on tag contents difficult. Of course del.icio.us is capable of developing a network based purely on the fact that an object was tagged…i.e it largely disregards the subjective categorization which becomes user level.
The power of last.fm lies in its ability to move beyond the assertion of a particular taxonomy or typology. The addition of groups, friends (as opposed to listening ‘neighbours’) and tags reintroduced the ‘subjective’ structuring of musical relations and for many this ‘social’ networking aspect has become the principle element of last.fm. This hasn’t yet threatened the diversity of playlists but has added a degree of peer influence to profile generation. Cliques tend to emerge and people increasingly move into ‘closed’ neighbourhoods with the ability to listen to another members stream etc.
that will do for now…