The tool sorts through 400 million tweets from 140 million active users every day and assigns a sentiment score to each tweet, either positive or negative. The scale runs on a 0-100 scale and is individual to each candidate, meaning they are not head-to-head. Both candidates can have a score over 50 and they can both lose or gain points on the same day.
Not only can the algorithm tell if a tweet is positive or negative, it can determine the severity of the sentiment. It is not without error, of course. For instance, it can't take into consideration any sort of sarcasm. That being said, the algorithm still agrees with the sentiment of a randomly selected human 90 percent of the time over 300,000 tweets according to Topsy's chief scientist Rishab Ghosh. That's not a bad record.
Twitter will release the results daily on the Political Index site. It already has a poll from May 1, 2012 through July 30, 2012 with a sweet line graph that shows some drastic changes. You can even track some of the changes with events, for example: May 10, 2012 was Obama's highest score of 74. That was the same day he attended a fundraising dinner at George Clooney's home. Chances are there was a lot of chatter about him that day.
It's great that a tool like this has come around to fill in for the traditional methods of opinion polling. By simply searching out generic tweets, there are no restrictions on what a person might say. They are not confined to simple black and white answers of a phone poll, and we get a lot more of a cross sample with the millions of Twitter users. While this may not be a direct poll, it certainly helps to see how the Twitterverse feels about the candidates, and it will be even more crucial as the November date draws near. That is, if these kids will actually go out and vote!
[]
Get more GeekTech: - - |