Mining Mood Swings on the Real-Time Web

Many companies are turning to social-media sites to gauge the success of a new product and service. The latest activity on Facebook, Twitter, and countless other sites can reveal the public’s current mood toward a new film, gadget, or celebrity, and analytics services are springing up to help companies keep track. Social-media analytics startup Viralheat, based in San Jose, CA, is now offering free, real-time access to the data it is collecting on attitudes toward particular topics or products. One of the first customers for this new service–called Social Trends–is ESPN, which plans to use Social Trends to show live popularity rankings for different NFL teams.`

Viralheat uses natural-language processing and machine learning to sift through Twitter, Facebook fan pages, viral video sites, and Google Buzz posts to determine the Web’s collective sentiment toward everything from popular browsers to Pepsi to Steve Jobs. The company sells its data and analytics service for a monthly fee, but CEO Raj Kadam says that Social Trends will provide a free way to people to access data the company is already collecting. When a paying customer asks Viralheat to track a particular term, they have the option to share that information publicly. Kadam says that about 70 percent of users agree to share this information.

Social Trends uses this information to provide a widget that can be embedded on a blog or website showing the sentiment around particular terms. These widgets stay connected to Viralheat’s data stores through an application programming interface (API) and are updated as the company collects more information. Viralheat believes the tool will be particularly useful for news sites wanting up-to-date infographics and for bloggers who want to track trends.

Anyone can create a Social Trends account and then search for terms they’d like to follow, although the company doesn’t have data for every possible term. The system lets users create charts tracking a single term or comparing several terms. Kadam says that Viralheat is able to open up live connections to its data because its infrastructure can handle working with large amounts of information. Viralheat custom-built its software and hardware and optimized it for the analysis it needed to do. For example, it created a Web crawler that can sift through data on the Web and manipulate it as it is collected.

Kadam says his company isn’t worried that its free offerings will decrease the number of paying customers. Social Trends widgets only offer a snapshot of the data that paying customers get access to (72 metrics instead of just five metrics), he says.

Viralheat is not the only company offering to mine Web users’ sentiments toward particular topics or companies. Alec Go, a Stanford University graduate student who created the “Twitter Sentiment” analysis tool, says there are dozens of sites offering such services. But he notes that many commercial analysis packages are closed off from public access.

Experts agree that sentiment-analysis tools are becoming increasingly significant as companies try to stay on top of the discussions happening across the Web. “Companies have a love-hate relationship with social media,” says Ed Chi, who is area manager for the Palo Alto Research Center’s Augmented Social Cognition team. These companies recognize that social media can spread a message faster than anything else, he says, but they’re also aware that it can easily get out of control.

Chi believes that eventually companies will need to track sentiment as part of a comprehensive public-relations effort. Future platforms could classify topics being discussed, suggest possible responses, and analyze a company’s message to determine how likely it is to go viral. “Sentiment analysis will be a component of a much larger dashboard,” Chi says.