The story goes back to 2008. Canadian musician Dave Caroll was travelling by United Airlines from Halifax to Omaha. His guitar got damaged in the airline’s custody, and his attempts to get compensation were thwarted by an insensitive staff. Frustrated, he penned a song, filmed it and put it on YouTube.

“I should have flown with someone else, or gone by car, ’cause United breaks guitars,” sang Caroll. The video did not just get millions of hits, it had a viral effect. Soon, all social media sites were agog with messages from people angry with United. The airline, facing the wrath of citizen journalism, paid up compensation to the musician to stem the company’s rapidly plummeting image.

Thus, the story of Dave Caroll became a textbook example for companies and publicity managers to take online responses very seriously. Globally companies began realising the importance of ‘perception monitoring’ and internalising the feedback to stay ahead of competitors.

But how does a company monitor public perception? Not merely through websites, emails, advertisements or public relation exercises, which do not facilitate a two-way communication with target groups. As more and more netizens bank on social media, rather than on websites or personal emails, to air their views, one thing that became clear is that cyberspace is the forum where brands can be built or destroyed.

This concept of ‘perception monitoring’ is the business of Kannan Raghavan and Ravi Puli, two NRI youngsters whose company Geomeme developed a piece of software that would crawl all over the cyberspace and track comments, positive, negative or neutral, being made on their clients. The backend operations of the U.S.-based company (www.geomemeonline.com) take place in Hyderabad, in southern India.

“What we do for our clients is slightly complicated. The software we developed checks for the keywords related to client and classifies them into positive, negative, neutral and mixed categories. Weekly and monthly reports are generated,” says Mr. Raghavan.

The software essentially uses the concept of ‘Federated Search’ (piggy-riding on popular search engines) and the links so gathered pass through hybrid auditing to arrive at sentiment analysis of a product or a company. Hybrid auditing means the captured web pages are checked by NLP (Natural Language Processing) algorithms for sentiment analysis, and later by social analysts.

“The NLP algorithms are statistical programmes which look out for expressions or phrases about our clients. But their accuracy could be only 40-45 per cent as the programmes cannot understand the context in which certain words are used. Hence we have social analysts who read every web page to determine the sentiment attached to a word. It is then the feedback is classified into good or bad or neutral. This enables us to increase the accuracy to nearly 90 per cent,” says Mr. Puli.

The social analysts function from their Hyderabad office. The human intervention also ensures tracking discussion forums and providing an avenue for companies to talk to netizens for getting opinions on a product.

Source:-http://www.thehindu.com/sci-tech/internet/article1138075.ece