Social search @ BayCHI
Last night's BayCHI seminar was a panel discussion by the heads of 5 hi-tech companies involved in new ways of finding information. In an unusual move, rather than having each speaker's bio read out, we were just shown the top-ranked Delicious tags for their companies. Netflix's Neil Hunt spoke about the Long Tail and how it forms the core of the competitive advantage that Netflix has over Blockbuster by letting them suggest cheaper indie movies to customers, which holds costs down.
Live365's David Porter began by describing the core model of his business: they make it very easy for anybody to create their own radio station, which leads to lots of viral marketing by DJ wannabes who convince all their friends to listen in. He talked about the 4 ways in which people find stations on his website: typing in an artist will list all stations that have that artist in regular rotation; you can browse by genre; people can bookmark their favourite stations; the system makes suggestions based on what users with similar tastes enjoy that you don't.
Pandora's Tom Conrad spoke about how the Music Genome Project spent countless hours deconstructing songs into hundreds of objective quantitative attributes to form a giant database that can now find similar songs across genres, time periods, etc. Since most people spend much more time listening to radio than their own music, Pandora lets them find new artists that sound similar to what they already like. Despite the obvious propensity for pigeonholing, he assure us that it exposes listeners to more new music than they would ever hear otherwise.
Joshua Schachter of Delicious revealed how the service stemmed from a previous incarnation that lacked multi-user support and existed solely for people to organize their own bookmarks and share them with a few select friends. Delicious was born when he discovered that there were thousands of people subscribed to his personal bookmarks! The advantage of this heritage is that the service overcame the initial inertia of most social networking systems by offering them immediate utility until the network effect could take over.
Digg's Kevin Rose demonstrated some features of the service of which I'd been hitherto ignorant: users can see what their friends thought of links; current stories can be visualized in a cloud where the popularity of each is indicated by font size; the Spy lets users watch as stories are rated in real time; a map that graphed users VS stories; a graph of people's friends that indicated what interests were shared between them. He vaguely alluded to taking more advantage of the social networking data in future.
The first topic of discussion was kick-starting the database with expert information. Apparently AZ seeds heir DB with book reviews to pick up information about more obscure titles. NetFlix uses metamoderation of users who categorize movies by comparing them to other movies. It was pointed out that transaction history lacks judgement info so users need to be given a very convenient way of rating their experience.
The next topic was about getting users to rely on suggestions made by the system. In a study by Pandora, people had no idea how interactively their ratings were being used so they added wordy explanations to the process. And Digg got much more user activity when they made it easy to rate stuff.
When somebody asked about Last.fm, Tom revealed that a Last.fm plus pandora mashup is in the works. The night ended with a brief discussion about accuracy in reporting. Both Digg and Delicious take the view that the people should be able to flag inaccurate stories themselves instead of relying on journalistic integrity.
Live365's David Porter began by describing the core model of his business: they make it very easy for anybody to create their own radio station, which leads to lots of viral marketing by DJ wannabes who convince all their friends to listen in. He talked about the 4 ways in which people find stations on his website: typing in an artist will list all stations that have that artist in regular rotation; you can browse by genre; people can bookmark their favourite stations; the system makes suggestions based on what users with similar tastes enjoy that you don't.
Pandora's Tom Conrad spoke about how the Music Genome Project spent countless hours deconstructing songs into hundreds of objective quantitative attributes to form a giant database that can now find similar songs across genres, time periods, etc. Since most people spend much more time listening to radio than their own music, Pandora lets them find new artists that sound similar to what they already like. Despite the obvious propensity for pigeonholing, he assure us that it exposes listeners to more new music than they would ever hear otherwise.
Joshua Schachter of Delicious revealed how the service stemmed from a previous incarnation that lacked multi-user support and existed solely for people to organize their own bookmarks and share them with a few select friends. Delicious was born when he discovered that there were thousands of people subscribed to his personal bookmarks! The advantage of this heritage is that the service overcame the initial inertia of most social networking systems by offering them immediate utility until the network effect could take over.
Digg's Kevin Rose demonstrated some features of the service of which I'd been hitherto ignorant: users can see what their friends thought of links; current stories can be visualized in a cloud where the popularity of each is indicated by font size; the Spy lets users watch as stories are rated in real time; a map that graphed users VS stories; a graph of people's friends that indicated what interests were shared between them. He vaguely alluded to taking more advantage of the social networking data in future.
The first topic of discussion was kick-starting the database with expert information. Apparently AZ seeds heir DB with book reviews to pick up information about more obscure titles. NetFlix uses metamoderation of users who categorize movies by comparing them to other movies. It was pointed out that transaction history lacks judgement info so users need to be given a very convenient way of rating their experience.
The next topic was about getting users to rely on suggestions made by the system. In a study by Pandora, people had no idea how interactively their ratings were being used so they added wordy explanations to the process. And Digg got much more user activity when they made it easy to rate stuff.
When somebody asked about Last.fm, Tom revealed that a Last.fm plus pandora mashup is in the works. The night ended with a brief discussion about accuracy in reporting. Both Digg and Delicious take the view that the people should be able to flag inaccurate stories themselves instead of relying on journalistic integrity.