It was almost exactly a year ago when OneRiot launched its realtime search engine. At the time, the playing field was much different. “Realtime” was just emerging as a hot buzz word, and Twitter had about half of the features that it has now. Facebook had just started going realtime, and I’d argue that FriendFeed was still the actual king of realtime (obviously, this was before Facebook bought them). A lot has changed in a year.
Today, OneRiot is rolling out a major revamp of both its site and some of its underlying architecture. The site itself looks much nicer and is better organized (sort of like a realtime Techmeme, in a way), but OneRiot’s site is just a fraction of what they do. That’s why the underlying architecture elements are much more interesting. In particular, OneRiot is today launching a new Trending Topics Engine, which it says is the fastest and most complete way to find the best things being shared on the web.
That’s a pretty bold statement given the competition. But OneRiot believes it has them all beat because of their new algorithms which hunt down items before they’re actually popular. Obviously, how this works is all technical, but basically they take a look at various streams of data from sources including Twitter, Facebook, MySpace, and Digg (Twitter is just a fraction of their overall data, they say). They then look for certain key phrases including things like “interesting” and “emerging” and flag the links attached to those elements to be “Trending Topic Candidates.”
From here they assign weights to each of these items, and filter out the porn and the spam. Everything is then clustered together to make the data more readable. A lot of this is done through semantic calculation.
There are a few steps from here, but you get the picture. OneRiot’s blog has a full rundown if you want to know more.
While the obvious use for this is to allow people to find hot items on the web quicker, there are actually a few other reasons this speed is important.
One is helping content-producing websites be able to know what stories to write about before they’re actually popular (but well on their way). OneRiot will be licensing a feed of this hot data to certain sites for this purpose (they won’t say which sites). So those sites that comb hot Google search results and write about those in the hope of ranking highly in hot searches may have found a new best friend.
Another use for this new OneRiot Trending Topic Engine is to better target trending ads. One of OneRiot’s main products is RiotWise, its realtime ad placement service. If OneRiot is better at knowing what will be popular before it gets popular, obviously, this can only help these ads. And it’s a way to continue differentiating its product from the ad product Twitter recently unveiled.
While Twitter ads allow brands to serve up certain tweets as ads at their choosing, OneRiot’s method automatically serves up these ads from brands to make sure they catch on to hot topics as quickly as possible. Again, this new Trending Topic data can only help that.
The final way this new engine can help is with search results. While there are no shortage of services offering realtime search now (including Google, Bing, etc), the real key is knowing when a specific topic is popular enough to warrant it showing up. With this new data, OneRiot says its partners will be able to beat their competition to showing relevant realtime results.
OneRiot actually built a display at their headquarters which shows their trending topics versus how quickly these topics show up on Bing, Google, Yahoo, and Twitter. I have to believe they wouldn’t have told me about such a display if they didn’t constantly beat their competition on it.