I loved Robert Scoble’s post on reverse engineering TechMeme. His description of how it works sounds right. Scoble describes something very similar to a genetic algorithm, which I’ve talked about in this context before. There is a set of sites that are the “fabric” of TechMeme. What TechMeme tries to do is measure how many of these sites are talking about the same thing. The more fabric sites are taking about a topic, the more it rises to the top. This fabric concept is TechMeme’s equivalent of Google’s Page Rank.
This is like letting each meme written about in the fabric be a test organism. The sites that wrote about the meme create a fitness function (along with other factors we’ll see in a moment). The more sites that write, the higher the fitness score, and hence that meme is a more “successful” organism.
There are other factors that TechMeme uses that are interesting. Scoble’s announcement of his new son got there because the fabric liked it and the words Scoble used mimicked the launch of a new technology, which is precisely what TechMeme is aimed at tracking, although Gabe Rivera says the system was mistaken to have flagged a baby as such. There are other such tweeks, such as taking a view that a longer story is more authoritative than a short story.
The discussion of what TechMeme doesn’t consider or what look like weaknesses was particularly interesting to me:
- It doesn’t consider video.
- It doesn’t consider competing non-blog aggregators like Digg, Reddit, or YCombinator.
- It isn’t very diverse. If you aren’t in the “fabric” of A-List bloggers (Scoble guesses 2000-5000 bloggers), you aren’t considered. An A-List can come along much later and you’ll be scooped if you aren’t in the fabric or unless a lot of people in the fabric link to you.
- Story length matters a lot. Too short or too long and you don’t get on TechMeme.
- TechMeme favors stories about itself.
- It doesn’t do foreign language sites.
- Stories about small technologies don’t get considered.
The overall picture reflects some interesting biases. The fabric represents a “hidden leaderboard” that is pretty well hard-wired into the system. Until that group picks up a story, it can’t move onto TechMeme. That’s pretty directly the “echo chamber” effect that some like Todd Cochrane have complained about. It’s basically a system that tells you what the top bloggers (as defined by TechMeme’s fabric) are talking about. The question to ask yourself is how you value that?
The top bloggers are not exactly hard to find. They talk about each other constantly for one thing. They have large authority on Technorati. Are you reading them already? Do you want them filtered for you by TechMeme so you only see things more than one of them talks about? Are you sure the top bloggers are talking about everything you want to know about?
I like TechMeme, but I like it more as an analytical tool than a news finding tool. Scoble reads 900 blogs (?!??), while I only read about 160. Nevertheless, I make it a point to read TechMeme last, and I am rarely ever surprised. Instead, I get to see which of the stories I just read from my other sources popped up there. In other words, it isn’t a meme discovery service for me, it’s a meme weighting service.