CNET también está disponible en español.

Ir a español

Don't show this again


Twitority refines Twitter search--or kills it

An experiment to improve Twitter's search results ignites a debate among top bloggers.

Twitority, the new alternative to Twitter's own search engine, is a brilliant idea. Or an incredibly stupid one. It depends on what you are looking for.

Over the weekend after Christmas, Seemic CEO Loic Le Meur put up a blog post saying Twitter's search engine is insufficiently discriminating. If you're looking for the best Twitter posts on a topic, he says, you want results from the most authoritative people to be ranked higher than Tweets from the rabble. Le Meur proposed a search engine that ranked Twitter posts based in part on the number of followers the poster had. He equated this, roughly, with the influence of the poster and thus the importance of the post. And then Jon Wheatly went and built it.

Respect my authority?

A kerfuffle ensued, which you can read about on TechCrunch, Scobleizer, and on Le Meur's blog again. The upshot is that there are people who maintain that popularity does not equal authority. They're right, of course. But that doesn't mean that Twitter's search engine, as it is now, has enough horsepower for today's users.

Search is complicated. Google's success in search is based on its initially clever twist on ranking search results, and on the continual evolution of its algorithms. Google shows us that recency matters, authority matters, popularity matters, and content matters too. Twitter Search, in contrast, is a blunt instrument--it only ranks recency. Twitority is more clever, and it shows us that Twitter Search needs to evolve. But it's far from the perfect Twitter search solution.

To illustrate, let me give two examples of Twitter searches, one where an authority-influenced algorithm would be very useful, and one where it most certainly would not.

Say I wanted to read the Twitter posts coming out of a conference (Le Meur's test case). In that case, using the conference's name as a search query might yield all sorts of useless results from people commenting on posts made by those who are actually at the event. Now, popularity of a poster does not equate to authority on a topic, as many bloggers have correctly observed. But having at least some kind of algorithm on top of recency could be a help if you're looking for the most relevant posts on a popular topic.

Twitter Search to the rescue.

And here's my counter-example. Last week, I wanted to know the driving conditions on the Siskiyou pass, the stretch of Interstate 5 that connects California and Oregon. Local Web sites said, "chains required." But they always say that in winter. A Twitter search on "Siskiyou" yielded better information--posts on road conditions from people who had just made the trip. Did popularity matter? Not one whit. Authority on these posts was earned by being there recently.

I bet that if Google's search engineers took a stab at indexing Twitter posts they'd quickly realize that on some topics, popularity matters. On some, it doesn't. On others, authority comes from elsewhere: Being first, being linked to, being different. Like I said, search is complicated. Twitority is a cool little experiment, and a good illustration that Twitter search could become much more useful than it is today.

Related: 10000 followers won't get you a free latte, on Live Digitally.