Attention profiling is a technique for syndicating learned user attention data between web services to help highlight interesting content and filter out uninteresting items. It works by tracking what users do on sites like Digg, last.fm, and Delicious and turning that into usable attention data that can then be syndicated between sites. Open source frameworks like Midgard and Apache Mahout make it easy for services to support attention profiling and learn from user behavior to better serve users while respecting their privacy.
16. How to actually do it?
Midgard Apache Mahout
• LAMP framework • Java and Hadoop
• Attention profiling part of the • Industrial-strength machine
API learning
• Good for simple services • Cloud computing
• Out-of-the-box APML • Less out-of-the-box
17. There, problem solved
Your web service learns, reduces
infoglut, and serves the user better
18. Play nice
Let user know what is being tracked
Give user their own attention data
Allow history removal
19. Wrap-up
Attention profiling can make your service smarter
More and more APML sources are coming up
APML is easy to support via Open Source software
www.apml.org
www.dataportability.org
Henri Bergius http://bergie.iki.fi