I spotted a new XML button on a blog yesterday:
The button links to an APML file that describes a person's interests, in an XML dialect suitable for consumption by software. APML, which stands for Attention Profiling Markup Language, has a short-on-detail spec that wasn't easy to figure out. There's an example and a schema, but no description of each APML element and how it can be used.
The web applications Engagd and Dandelife support APML, so I joined them to see how they use APML to describe my interests. Like John Tropea, the blogger who led me to APML, I like the idea of pulling this kind of data out of sites like Amazon.Com and Netflix so you can use it elsewhere.
I publish an OPML reading list of the RSS feeds I'm currently reading. Engagd will analyze this file to determine the subjects most of interest, storing the results in APML. The site decided that I'm geeked about these topics:
<Concept key="google" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="news" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="college football" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="team preview" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="conference preview" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="performance" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="general" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="java" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="miscellaneous" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="news," value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="more from athlon sports" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="sun belt" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="feature" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="yahoo" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
<Concept key="game" value="0.50" updated="2007-08-31T10:52:45Z" from="profiler.engagd.com"/>
The key attribute lists my biggest interests, according to Engagd. Before there are any misunderstandings, my concerns over "performance" are not personal in nature. I'm interested in the performance of the North Texas Mean Green.
Key interests should change as my reading list changes, because Engagd keeps monitoring it. The site can use my APML data to recommend items from other feeds, producing a new filtered feed in RSS format. Here's some Digg stories Engagd thinks I might like. (Note: The feed doesn't validate because of a rank element that's not in a namespace and other issues.)
APML doesn't show much promise in Engagd's current recommendation engine, but it's a new app limited by the data I gave it: one reading list. APML's designed to hold information about all of the stuff people are interested in -- products bought, movies viewed, books read, web sites clicked, celebrities stalked, and so on -- that they decide to offer for public consumption.
Excessive sharing gives me the heebie jeebies -- does the world really need to know much time I spent reading about Jason and Sam's breakup on General Hospital? -- but I'm going to start publishing an APML file on Workbench and pay some attention to the project. I don't like the green icon that Engagd developer Chris Saad has suggested for APML data, so I'm creating my own:
To get my attention, click my monkey.
Credit: Attention Monkey is borrowed from the Tango Desktop Project and is available under the Creative Commons Attribution Share Alike license.
