Pretend, for a moment, that popularity can be quantified. Like, with numbers. The more social-media comments about something, the better. In practical terms, that means we care a lot more about American Idol and football than we do the Republican primaries.
Fast Company recently reported on Bluefin’s ability to group comments into positive and negative. Now they’re moving into advance sentiment analysis–not just if they like something or not, but that illusive why. “If positive-negative is one dimension, then excited-bored is another,” Bluefin CEO Deb Roy says. “You could, for example, hate something and be excited about it.” Besides getting actual people to agree on what counts as excited and what counts as bored, Bluefin has to, of course, teach its machine-learning algorithms to recognize the emotions too.
With a just-announced $12 million in series B funding led by Time Warner Investments, Bluefin can fund a lot of lessons. For a glimpse of what’s to come, Bluefin gave us an analysis of the most recent Republican debate. Their algorithms “figure out” actual topics in tweets–not just basic keywords. And they show what exactly people were talking about most: Ron Paul’s controversial newsletters, followed by the fact that Jon Huntsman can speak Mandarin (the meaning of that is left up to the reader to glean).
Still, the ability to organize social-media users’ complete conversations isn’t quite there yet. Now the company can, for instance, group the most common words people use on social media to talk about a specific Diet Coke commercial–and reveal the shows those people prefer, creating a sort of tv-show-Diet-Coke Venn Diagram. But to figure out why people are drawn to those things using something more nuanced than single words, you’d have to sift through and read the comments. “It’s still a manual process,” Roy says.
Teaching computers to do that for us will reveal “that there’s actually some main events that happened that drive the conversation,” Roy says, which is a boon to brands that want to get people talking not just about the right things but also in the right way. To get there, Bluefin engineers have to figure out how to get past the ambiguity of language in online comments, whose length limits and slang make them tough for computers to understand (“bad” can mean “good,” and sarcasm, the go-to dialect of social media, is tough to pin down). One solution, Roy says, is having the machines look at a person’s comment relative to what that person’s said in the past.
Aside from using its new cash for developing technology, Bluefin will also use it to expand its sales and clients services team. “Our primary motivation is to accelerate taking Bluefin Signals into the marketplace,” Roy says, and, at the same time, making sure Bluefin stays ahead: “Our system is as good as anything you’ll find in the analytics industry, but we think we can do a lot better.”
[Image: Flickr user Beacon Radio]