An Indian startup is using natural language processing tools to mine Twitter for news and events that will likely dominate the news cycle.
Frrole, a Punjabi word loosely translated as "serendipitous discovery", was created by Bangalore-based Amarpreet Kalkat whose algorithms identify Tweets relevant to individuals living in a particular city. For example, the Bangalore page would display a range of stories from different Twitter accounts, including a Time magazine story about how local police officers were using cardboard cutouts to curb reckless driving.
Kalkat launched the site in April 2012, and at the end of the year recruited co-founders Abishek Vaid and Nishith Sharma, to oversee the technical development and market the site, respectively.
Vaid, who quit his teaching job early last year to pursue an alternative career path, became involved in the business by chance when he reconnected with one of his senior classmates from IIITM Gwalior, an alum that had previously worked with Kalkat at technology company, Trilogy.
The two were studying an online course from Carnegie Mellon when the classmate asked if Vaid would be keen to go to Bangalore and build a product. "He completed the triangle," he recalled.
Vaid leveraged natural language processing research and libraries from the University of Washington, Carnegie Mellon, Lingpipe, and Twitter text-mining specialist, ARK.
Together, the trio saw site visitors grow to approximately 700,000 a month, and Sharma expects to breach the 1 million milestone in the coming months.
With 50 percent of the Indian company's users in the United States, he said Frrole is one of the few Indian startups that has developed a global software product based on Twitter.
"India has typically been a service-oriented country [supporting] clients, or recently what we've seen is that most startups focus on the. But there's no hardcore, technology-based product. So we're among the first," Sharma said.
Kalkat is currently in the United States talking to investors, and is also in discussions with Twitter about becoming a formal partner--which, if signed, would make Frrole the first Indian company to do so.
While most of the content are generated by existing news organizations, with its technology, the Indian startup aims to identify important news events the moment they are tweeted by first-hand observers and immediately promote them to the relevant audience.
One example occurred during the recent Delhi gang rape scandal where police arrested a woman protesting against the government's response in the case. She tweeted the event, which then was heavily retweeted--a key metric in Frrole's algorithms. This pushed the news to the front page of the Frrole site at least 15 minutes before it caught the attention of mainstream media.
In essence, the site aims to predict the news, Vaid said. "In our generation, news is no longer in the hands of journalists. Each of us is a source for generating news [and] Frrole is trying to capture this paradigm."