Information overload is becoming a major issue, and more sophisticated solutions will be needed to tackle it. With Midgard we’ve already taken some steps into this space by creating tools for calculating newsworthiness of stories in order to present only relevant ones to a user.

After being able to calculate objective interestingness of various data items, the next phase is to make the filters more subjective, more personal. Attention profiling is a tool for this. There are already some attention-based services on the web that work quite well. For example, Amazon is able to make quite accurate book recommendations to me, and the concerts suggested by Last.fm have mostly been interesting. But so far this all has happened in closed silos.

APML is an emerging standard for syndicating attention profiling information between web services and applications. While not many services support it directly, there are some third-party tools for gathering the data from various sites.

There are many places in Midgard where attention profiling could be used to provide better service to users. My plan for APML support is following:

Generating attention data:

• net.nemein.favorites: faving items generates positive attention for item's tags and categories, burying items generates negative attention for item's tags and categories
• net.nehmer.blog: tags of blog entries created by user generate positive attention
• net.nemein.bookmarks: tags of bookmarks generate positive attention

Using attention data:

• org.maemo.socialnews: use attention as additional modifier when calculating what to show
• org.routamc.photostream: create a new "most interesting" view of photos based on photo tags and user's attention

In order to make this all possible there should be an attention handling library that would provide easy APIs for attention calculations.<p style="text-align:right;"> Technorati Tags: apml, attention, infoglut</p>