A Sentiment Detector That Reads the Social Web for You
BY DAVID ZAXThu Oct 28, 2010
Visible Technologies, a service that helps companies track what’s said about them on the Web, unveiled a major upgrade today. Its new Google-inspired architecture better distinguishes the relevance and the sentiment of the mountains of social media conversations. It aims to help businesses make sense out of the cacophony of the social web, and clients Microsoft, FedEx, Dr Pepper Snapple Group, and others are all clients of the service.
Visible has been around for a while, and has plenty of competition (we recently wrote about Trendrr, for instance). But the latest iteration of Visible’s platform, called Visible Intelligence, has been built from the ground up. It has a scalable, Google-like architecture that can render tens of millions of blog posts, tweets, and forum entries within seconds.
The platform also has some new bells and whistles. There’s a sophisticated sentiment analysis algorithm, for instance. Teams of people from all over the country combed through blogs, forums, and tweets, scoring individual communications on a sentiment scale. Gradually, over a period of five years, these legions of social media scavengers nurtured a sentiment-detection algorithm–one that can distinguish between when someone says a product is “sick” in the positive sense a snowboarder would use, and when someone says it in the negative sense an outpatient would use.
Visible’s new structure also takes into account the fact that engaging with social media is an “enterprise-wide strategy,” not just limited to the PR department. Customer service, for instance, can benefit from Visible. Says Mike Spataro, VP of Enterprise Client Strategy: “Consumers today aren’t emailing or phoning as much–they’re now on Twitter complaining.”
If you gripe about FedEx in a public tweet, in theory, a customer service representative could come to you, rather than the other way around. It might seem a little creepy, but when it happens, Spataro says that “98% of consumers are surprised and delighted that someone is trying to help them, and don’t think it’s like Big Brother trying to steer them.”
Visible has plans to roll out new services for the platform every six weeks. In the pipeline, for instance, is a way for its users to distinguish between different kinds of influencers–a Twitter all-star who has a wide audience, for instance, versus a less-well-known but highly authoritative voice. “Imagine you’re a tech company, and what you want to do is reach people about your cloud services,” explains Visible Technologies CEO Kelly Pennock. “In that case, you care less about reach, and more about authority.”
Visible is also taking a cold, hard look at the sheer amount of data out there. It has tapped Enrico Montana, a Ph.D. in neurobiology from MIT, to tweak and hone its algorithms. He used to measure genome expression changes in organisms; now he tweaks the equally complicated organism of the social web. “As data sets get bigger and bigger, have to be more intelligent on processing side,” says Montana.
Ten billion tweets a year (according to one projection), not to mention blogs and forums, will amount to a whole lot of noise. Visible’s insight is not only how to pull the signal from that noise, but how to pull multiple signals, depending on the question you want to ask.
[Image credit: Flickr user williamcho]