Welcome to TasteBroker.org

TasteBroker is a set of experimental web services for generating and processing taste data. TasteBroker currently supplies two types of web services:

APML Generation Web Services

These web services provide an APML representation for a user's taste data based on data supplied from various web destinations.

APML from last.fm

If you are a last.fm user, you can get a representation of your music listening taste, in the form of an APML file. Simply use the webservice:

http://aura.darkstar.sunlabs.com/apml/music/Your_last.fm_name

Some examples:

Try it yourself:

http://aura.darkstar.sunlabs.com/apml/last.fm/

Note that it can take a while to generate your APML file since it requires a number of queries to last.fm to collect your taste data. We try to cache as much info as we can to make things quick, but if you have eclectic tastes, you might have to wait a minute or so.

APML from Pandora

If you are a Pandora user, you can get a representation of your music listening taste, in the form of an APML file. Simply use the webservice:

http://aura.darkstar.sunlabs.com/apml/pandora/Your_Pandora_name

Some examples:

Try it yourself:

http://aura.darkstar.sunlabs.com/apml/pandora/

Note that it can take a while to generate your APML file since it requires a number of queries to Pandora and last.fm to collect your taste data. We try to cache as much info as we can to make things quick, but if you have eclectic tastes, you might have to wait a minute or so.

APML from del.icio.us

If you are a del.icio.us user, you can get a representation of your web browsing taste, in the form of an APML file. Simply use the webservice:

http://aura.darkstar.sunlabs.com/apml/web/Your_delicious_name

Some examples:

Try it yourself:

http://aura.darkstar.sunlabs.com/apml/web/

Recommendation Web Service

This is an experimental web service that provides music recommendations based upon the APML representation of your taste. This web service call returns an APML file with a 'Music-Recommendations' section that lists artist recommendations based upon your listening habits.

Some examples:

The algorithms used to create recommendations are described in Tagomendations: Making Recommendations Transparent.

The web service takes a number of parameters:

Parameter Default Description
apmlURL None The url of the input apml file
outputFormat apml The output format. Currently supported formats are: 'apml'. Coming soon: xspf
alg default The recommendation algorithm to use. Current algorithms are
  • imp1 - recommendations based upon your implicit tastes
  • exp1 - recommendations based upon your explicit listening behavior
  • exp2 - alternative recommendations based upon your explicit listening behavior
  • exp3 - recommendations based on your explicit tastes (a faster version)
  • default - use a well-rounded recommendation algorithm
type artist The type of recommendations to generate. Currently supported types are: 'artist'. Coming soon: track
num 30 The number of recommendations to generate.
profile defaultprofile The APML profile to use when generating recommendations. If none is specified, the defaultprofile found in the APML is used.

Some more examples

Recommendations based upon BBC Radio shows:

Last.fm users try it yourself:

http://aura.darkstar.sunlabs.com/apml/last.fm/

If you are a last.fm user, enter your last.fm user name, and an APML with recommendations will be created for you.

Pandora users try it yourself:

http://aura.darkstar.sunlabs.com/apml/pandora/

If you are a Pandora user, enter your Pandora user name, and an APML with recommendations will be created for you.

Recommendations based on APML from the BBC:

What is this all about?

This is an experiment to explore web services that produce and consume APML.

We can combine this APML generator with other services that process APML. For instance Tagurself provides some Javascript that will turn an APML file into a tag cloud. For example, here's a couple of tag clouds generated from my taste data:

Paul's web browsing interests:
Paul's Pandora music interests:

There are also some usage stats.

History


TasteBroker.org v.73 - Powered by APML, Cluztr, Del.icio.us, Last.fm, Pandora, tagurself.com and Sun Microsystems inc.
Send comments/feedback or complaints to Paul.Lamere@sun.com
Don't forget to read Duke Listens!