Bath, 29 April 2015 – From now through to election day, political consultants, journalists, and news junkies can watch how parties in the upcoming UK elections are faring in social media in real time using Brandango, a sophisticated sentiment analysis tool developed by Tivarri Limited in Bath.
"Watching the social media conversation about the UK parties as it happens is easy, fun, and available to anyone who visits http://opinion.expert" said Tivarri founder and CEO, Simon Ponsford. "Whilst the free version is not as rich in features as the full Brandango platform is, it does track and compare sentiment in real time for each of: Liberal Democrats, Conservatives, Labour Party, UKIP, Green Party, SNP and Plaid Cymru. By hovering the cursor over the graphs, the user can see point-in-time results in fifteen-minute intervals on the daily chart and four-hour intervals on the weekly chart. Anyone with an interest in politics can use it to gain significant insights into the perceptions of the various candidates from now until Election Day."
The web page features a live Twitter feed and displays the percentage of comments made about each party in the past 24 hours. It also shows the percentage of comments made about each party by gender. For example, women typically make about 28 percent of all tweets about the Liberal Democrats (the lowest female percentage), but they make about 38 percent of the comments about Conservatives (the highest female percentage). The page offers a word cloud and comment counter by location that users can manipulate to see the locations of comments about each party.
Brandango is immune to organized efforts to manipulate the social media, also known as "astroturfing," according to Ponsford, who oversaw the development of the software. Numerous businesses already use Brandango to monitor market perceptions, identify business opportunities, and prevent negative comments from turning into a crisis. "The size of our data sets and our software's 'secret sauce' detects and deflects organized attempts to influence social media," said Ponsford. "What you see in Brandango are the perceptions of real people in real time."
The engine behind Brandango's accuracy is "big data," which Wikipedia defines this as "any collection of data sets so large or complex that it becomes difficult to process using traditional data processing applications." Brandango is able to process millions of social media comments a day, something only a handful of companies and universities can do. But according to Ponsford, receiving and saving the data is the most straightforward aspect of the challenge. "Understanding sentiment, picking up key phrases, checking the gender, geographic location, measuring mentions and determining the influence of the person's comment in real time is a much more complex computing challenge," he said.
Speaker influence is critical. A tweet by a person with 100 followers cannot be given the same weight as someone with 100,000 followers. But influence is easier to gauge than the language the speaker uses. "Because people use sarcasm and irony to express themselves, assessing whether a post is positive or negative can be difficult," said Ponsford. The Brandango sentiment analysis engine has been repeatedly tested against human analysis of the same data, using large groups of students, and refined to make sure it is accurate. It also has been tested against other sentiment analysis engines. "Our tests reveal that Brandango has the highest accuracy in the industry," said Ponsford.
Brandango is one of the few platforms that can perform sentiment analysis "on the fly". The majority of companies offering sentiment analysis tools rely upon batch processing of data that is collected once per week or even once per month. By the time the analysis is complete, the data is days or even weeks old, and the organisation receiving the report is not able to respond to positive or negative trends in a timely fashion. By processing streaming data, Brandango offers real-time analysis. "You can literally watch trends emerge," said Ponsford.
To try the Brandango sentiment analysis tool, go to: http://opinion.expert
To find out more about Brandango http://brandango.net