Let’s continue talking about feed reading. Last week I described the five phases of feed reading I had observed. Or actually the first four phases are observations, while the fifth phase is purely speculation.
I don’t agree that that’s phase five.
Why? Cause I keep meeting interesting people I want to have conversations with and who aren’t in Memeorandum (or who won’t be in the tabloids).
I have to admit that I did not think a lot about the fifth phase when I wrote it down. It was simple the first idea that came to my mind, and I jotted it down. Raw blogging…
But after reading the reactions, I thought it oer again. What exactly do I expect? How would I like to receive my news?
The tabloid model and Memeorandum do a nice job in selecting, but there is one thing neither of them covers: they are both excellent in selecting the top stories which everybody likes, but fail to select the top stories I like. There is no personalisation, and feed reading has everything to do with a personal experience.
What we really need is ‘clever’ feed readers, applications that know what I like and what I am not interested in.
My first idea is to use a Bayesian filter, which is already succesfully used to identify spam and to route helpdesk requests, on a big set of RSS feeds and have it select the most interesting items. I rarely have very original ideas, and there are many people spending more time on feed readers, so somebody must already have had the same idea, and maybe even already worked it out. A little bit of searching lead me to 0xDECAFBAD, who tried it out and found that it did not work so well.
I know that the idea is still new, and that this is just one test conducted by one person, but it is clear that more work has to be done in this field. For now I start playing with the new TailRank service, which partly solves my problem according to all those people who have written about it the last days.