Tim Ireland (and pals) have been working on Blogcode, a blog recommendations service. Rather than go down the long and difficult Web 2.0 road of autodiscovery, comparing mutual links, funky aggregation etc., it goes for a simpler method: Users merely “code” a blog – i.e. rate it on different metrics, and the system aggregates the metrics together, to find out what blogs are compatible.

I’m not sure how well it will take off – this kind of recommendation site only works if adding & recording your opinions is made as effortless as possible – which is why e.g. works so well. What they needed to go with it would be, say a Firefox extension (or even a Greasemonkey script) that would say, pop up infrequently from time to time when you visit a site that looked like a blog (e.g. autodetection of RSS links, or meta generator tags that say Blogger/Wordpress/Movable Type etc.) It would ask a single question about a single metric (“How would you rate this blog – left or right wing?”, “Is this blogger serious or light-hearted?”) and then bugger off (it could always give you the option to answer more questions, if you had a few idle moments). It would take longer for the data to build up, but wouldn’t need us to spend hours going through all the blogs on our blogrolls before we got bored and wristache from dragging those scroll bar controls.

Also, I’ve noticed one big fat bug already – it’s possible to have two separate codings for the same site (I started one for after being told no code existed for it, and then it turned out someone else had also done one). Still, it makes for amusing comparisons between what I think of my site and what the other indexer thought.

There are still some amusing glitches (I’m 71.51% compatible with Harry’s Place? Reminder to self – must write more anti-war polemics) – and I’m not sure what mechanisms they have in place to stop deliberate spoofings (e.g. lots of people giving spurious metadata), or the forever-problematic blogspam. And it would be cool if they released the raw data (with the contributors’ identities anonymised) for the ratings of different blogs; it would allow other developers the ability to perform some nice open-source hackery – this human-generated data, coupled together with computer-generated data from the Technorati API, could make for some really cool visualisations & mappings of the blogosphere.

Right, getting tired. Interesting stuff though – may ramble a bit more on this later.

2 thoughts on “Blogcode

  1. Yeah, the double entry was something I noticed as well; I entered me and someone else entered me, at almost the same time. Best bit was we were only 71% similar. Told Tim, he got the entries merged.

    I, also, got my highest match with Harry’s Place at first, no idea why, now we’re much closer to blogs I actually read, which is good.

    ’tis a nice tool, I’ve found one or two I didn’t know about. OTOH, I alreayd have too many blogs to read…

  2. I have a hypothesis about the Harry’s Place similarity – the first entries were done by the site’s own contributors, who self-style themselves as liberal and left-wing; as time went on, a more realistic appraisal started coming in from readers of the blog, which is why there is no similar rating now (I see that amongst others, similarly-rated blogs to HP include Samizdata and Conservative Home…)

No new comments may be added.