April 07, 2007

Robot-Generated Sites Show Occasional Flaws

TechMeme, Megite, TailRank and other sites aim to organize the hot topics of the day, through tallying those news stories and blog posts most frequently linked to in a given day. And for about 99% of the cases, the sites do very well, even if they do tend to tilt in favor of the A-Listers a bit, meaning that if TechCrunch and I were to both post an article at the same time on the same topic for example, TechCrunch would take prominence, due to popularity, not necessarily the quality of writing. (See example from Saturday)

Today, while browsing Megite, a technology news site that aims to tell visitors "What's Happening Right Now", I was surprised to see that HP had agreed to buy Compaq for $25 billion. That's a huge acquisition to be sure, but one that happened six years ago. Amusingly, the story was accompanied by a photo of long since deposed HP CEO Carly Fiorina. The big news came not because it was "happening right now", but instead because a pair of blogs had linked to the years-old story, from CNET's News.com, in the context of new coverage around the computer company.


Click the Above Picture to See Megite's Flashback


Essentially, if enough top ranked bloggers opt to link to a story, regardless of its age, it looks like they could propel it to the top of Megite and other sites like it. And that's broken.

TechMeme hasn't been as loose, but I've seen the occasional oddity.

For example, on March 23rd, I had written that "The Apple TV Debate Is Upside Down", and as part of that note, I had linked to a comment from Robert Scoble, who said "Apple TV Rocks". Not surprisingly, Scoble was carried by TechMeme, and I was a related story. (See the screen cap from Noon on March 23rd)

But I had also gained a lot of traffic from MacSurfer and MyAppleMenu, two Mac aficionado sites who had found my comments interesting. When I wrote a follow-on post, saying "Welcome MacSurfer, MyAppleMenu Readers", that post somehow supplanted the previous article's placement in TechMeme. Somehow, the page generator believed that the second post trumped the first. (See the screen cap from 3 pm on March 23rd)

In theory, after my first article had been linked, I could have changed its title to something profane, or its content to gobbledy-gook, after having gotten the TechMeme pickup. So long as I retained the same linkage from the first piece, whatever was presented on TechMeme was up to me.

The practice of having algorithms define what's hot and what's not, exemplified by sites like these, and Google News, is not yet perfect, and continues to be refined. I expect we haven't seen the last of the quirks as they continue to develop.