By Sam Pauken

Last Wednesday, Twitter released a new feature on their website – the Twitter Political Index.  According to Buzzfeed, by analyzing 400 million tweets from 140 million active users, they are able to provide a daily assessment of how the Twitterverse feels about President Obama and Candidate Romney. 

In coordination with the parties’ polling companies The Mellman Group and North Star Opinion Research, this looks like the most serious effort to gauge voter sentiment via analysis of data on social networks this election cycle. 

400 million tweets is a massive amount of information to pour through. To put that in perspective, on Election Day in 2008 there were a total of 1.8 million tweets.  Today there are that many tweets in about six minutes.

Twitter is not usually in the business of analyzing the massive amount of content that gets posted to its site every day.  That’s where the computer program Topsy comes in. 

Topsy analyzes every single Tweet in real time, figures out if the tweet is related to Obama or Romney, and then determines the sentiment behind the tweet.   How is this possible?  Topsy says their technology has been tested on 30,000 tweets and the algorithm was able to accurately determine the sentiment of a tweet 90% of the time (compared to a human).

Topsy’s chief scientist Rishab Ghosh (@r2g2) believes that this solves one of the biggest challenges facing opinion pollsters – gathering data from the mobile-only individuals that they can’t reach.  That may be true – but as far as the Twitterverse being representative, it’s well known that Twitter users lean young, urban and most probably liberal (see here and here)

Ghosh is realistic about the application of this tool: 

“We are not actually are measuring whether people prefer Romney to Obama or not. It’s just “a broad sentiment measure of what’s going on.”

According to Adam Sharp from Twitter, “I don’t think we would claim at this point that it is predictive,” though “someone who makes predictions based on this data, along with other data is going to be more informed than someone using one data source.”

We tend to agree.  Technology is giving us access to all sorts of new data sources and each one is a new window for possible insights, without worrying about how “representative” it may or may not be. 

TechPresident feels differently.  They note that although the Topsy algorithm and a human being agree 90% of the time on the sentiment of a tweet, sentiment, even among person-to-person interactions, is often difficult to determine and subject to misunderstanding.  Thus, Topsy, being a computer program, is even more unreliable, and the index cannot be viewed as 90% accurate.  It is only in agreement with a person’s subjective interpretation 90% of the time, making it an unreliable tool to determine the Twitterverse’s true sentiment. 

In the end, the Twitter Political Index might be a fun way to get a sense for the Twitterverse’s feeling on the candidates. It’s too early to say that these data are an accurate representation of its views, let alone using this index as a gauge for the voting population at large.