Radhika Krishnan, a professor at IIIT-H, said that an old-style statistical analysis will not give a correct picture and added that use of sophisticated tools that are often used in the Language Technologies Research Centre (LTRC) lab can actually decipher the nuances between the same word used perhaps by different newspapers or political parties.
“Just the frequency of the appearance of a particular word or hashtag will not tell us anything about the associated sentiment. For a social scientist, context is everything,” she says adding that challenge lies in mapping out trends on digital media.
While an initial frequency analysis will narrow down on the major issues that are on the radar, a more sophisticated analysis will be done to understand what people are saying about those issues, says Krishnan. She feels that issues like Loan waivers, corruption, Ram Mandir, agricultural prices, among others could assume significance in the upcoming electoral battle.
“Even if these words do not appear on Twitter or appear peripherally, I’m going to see if there’s a change,” she said adding that region from where a tweet is tweeted is also important as a serious issue in one area might not be relevant in other.
While Krishnan is trying to analyse data, a visiting professor at IIIT H Ponnurangam Kumaraguru’s (PK) efforts revolve around unravelling the extent to which social media can be manipulated.
“A popular method of creating an impression on a viewer about a political candidate is by altering the candidate’s followers online. When the number of followers surges overnight for a particular candidate, most of these followers are unlikely to be real people,” says PK.
An expert in security and privacy in online social media, PK’s previous work in 2014 involved heavy analysis of poll-related Twitter data. He and his student’s analysis revealed that activity on Twitter peaked during important events related to elections.
He said that he is interested in building a tool for Twitter like the one that exists for an e-commerce site, which helps one make an informed choice on the product under consideration by categorizing the reviews as an original rating vs adjusted rating.
“The tool will basically show the number of followers vs adjusted number of followers implying that the adjusted number are BOT followers, he said.