´óÏó´«Ã½

is our latest prototype and a spin-off from our previous work on people’s music taste. It showcases some of the hottest bands and artists on the web, as identified by a number of independent sources, lets you listen to short clips and shows you where you can find that music on the ´óÏó´«Ã½.

We use data from a number of independent sources to determine what music is trending - the buzz about music on the internet, usually based on what’s being talked about, what’s being played, what’s being sold, what’s being written about and more. All our current sources have public APIs. From we use the , from we take their and from we create a combined chart from their .

We combine the source charts by creating an average position for each artist, giving each chart equal weight. This means that artists who appear in more than one chart are more likely to be higher in the combined chart. We display only the top 15 artists from each chart on the site, as individual sources and a chart. The combined and individual charts all have two view options, built from ´óÏó´«Ã½ data:

- This is the top 15 trending artists sorted by the number of plays that artist has had on the ´óÏó´«Ã½.

- This is the top 15 trending artists sliced by radio network, based on which network played that artist most, and then sorted by number of plays.

The refresh rate for each of these trend sources varies but we pull their data in every night then build our combined chart from the three sources. To import the "hottt" artist list from EchoNest we extended the library and the changes we made are available on GitHub . After this we try to match everything to IDs and link the artists to ´óÏó´«Ã½ Music pages and music clips from ´óÏó´«Ã½ programmes. Where the source data does not include a MusicBrainz ID we only have the artist name, so we use the and Ruby libraries to obtain the ID. But there is always the problem of how to disambiguate different artists with identical or similar names, such as Eagles, Eagles, and The Eagles, so in cases where we can’t uniquely identify an artist we don’t link to the ´óÏó´«Ã½ website or provide audio clips. Fortunately, this doesn’t occur too often.

All the ´óÏó´«Ã½ data that we use in the prototype (number of plays, links to /music pages etc.) is taken from the public data views on /music and /programmes. The music clips are taken from ´óÏó´«Ã½ broadcasts and for each artist we present a clip from the most recent of their songs played by ´óÏó´«Ã½ Radio, along with a link to the ´óÏó´«Ã½ show it was taken from.

Developing this prototype also allowed us to try some features from the new HTML 5 standard. For audio playback we use the jQuery plug-in, which uses the element in browsers that support MP3 format audio (such as Chrome and Safari) or a Flash player in other browsers.

It’s a pretty straightforward prototype that looks useful to us and has an obvious purpose, showing you what the buzz is in music on the internet and showcasing the ´óÏó´«Ã½ content around that. Ultimately we have been thinking about doing more research into trending and data mining on the internet. If we do then it seems that we, as the ´óÏó´«Ã½, should be open and transparent about how we identify trends and in this particular case, we should focus on live, UK and new music.