Calculating news item relevance
I’ve started working on a new Social News section for maemo.org. The idea of this area is to provide a centralized view on what is happening at the moment in the maemo community.
Every day brings dozens of maemo-related posts via various channels, and keeping up-to-date with them requires a lot of time. The new social news section aims to fix this by providing a somewhat Digg-like news aggregator that will bring only the most interesting items to the top.
Interestingly, a new service called AideRSS went live today with quite much publicity. AideRSS is a new breed of RSS aggregator that uses various metrics to determine the relevancy of new items. This is what AideRSS says about most interesting stuff now on Planet Maemo:
While I don’t have access to their secret sauce, using a bit similar metrics I get quite similar results as well:
The way the new org.maemo.socialnews score calculator works is that it looks for number of votes or links from various sources, gives them configurable weight, and then builds a relevancy value out of that. This seems to work quite well, although I guess I will end up tuning it quite a bit when we start syndicating larger amounts of data.
In any case, the next challenge is to combine the relevancy data of items and their tagging/categorization to build a newspaper-like page. Actually, feeding this data to a proper newspaper generator could make interesting results as well.