State of the Attention Problem

Technorati's David Sifry has posted the latest of his regular updates providing a bunch of stats and analysis on the blogosphere's development.

The bottom line is that the number of blogs created, according to Technorati, continues to double about every 5.5 months. Spam blogs (splogs) account for around 9% of all blogs.

The data is showing once again that the blogosphere's content creation activity is a largely event-driven affair - real world dramas correlate with the number of posts published.


State of the Attention Problem

This corner of the internet is the live web. The blogosphere is annotating and analyzing the real world in real time. And at a furious rate - currently 50,000 posts are published each hour.

So we're creating more and more content, but how do we discover what's really of interest to us?  Technorati's data underscores the Attention problem. As David notes:

"At that rate, it is literally impossible to read everything that is relevant to an issue or subject, and a new challenge has presented itself - how to make sense out of this monstrous conversation, and how to find the most interesting and authoritative information out there."

I wouldn't say it is a new challenge that has presented itself, but I would agree that it is an increasing challenge that is worth tackling.

The live web discovery problem is different type of discovery problem to that the traditional search engine space has been trying to solve. Companies such as Technorati, Icerocket, PubSub, Memeorandum, Tailrank, Digg, FeedDemon, Rojo, and Bloglines and many other start ups that have cropped up in the last couple years recognize this and are helping us navigate the torrent.

However, in my view, what's missing from the current generation of the aggregators, feedreaders and live web discovery engines is the ability to scope these services against my attention data. Some of these services provide tag and keyword RSS search subscriptions and have some personalization features.  These are steps in the right direction, but we've got a long way to go.

Attention data, as I've tried to define previously, is a larger and more dynamic data-set than keywords or tags or clickstreams and it could and should be leveraged to provide significantly more relevant experiences.  I'm convinced of this. Watch this Attention space.


Tags: , , , ,

Comments (5)
  1. Waseem Sadiq says:

    Perhaps domain specific meme trackers are a part of the solution? I have written a piece on this over at my blog:

  2. OPML has got to be the way for memetracker sites like mine ( and the others you mention above to *begin* to filter the mad foaming river of blog data for the user.

    My view is: start with OPML; overlay tags; monitor/collect user interest over time (what content is liked/disliked); tune content based on interest.

    Throughout this, the memetracker should suggest new tags/links/content that might interest the user but isn’t quite within their sphere of attention. Tangents, if you like.

  3. The word ‘memetracking’ or (‘meme tracker’) has been used to describe services such as Memeorandum, Megit,…

  4. Joshua Porter gave a talk recently ‘Web 2.0 – Leveraging the Network’. He has published the slide deck…

Comments are closed.

Skip to main content