Over the past week, there have been three interesting events involving President Barack Obama and Governor Mitt Romney, two US Presidential debates and the Alfred E. Smith Memorial Foundation Dinner. One of our Data Scientists, Henning Gabriel, created an app for our brand-new Analytic App Market™ called Twitter Election Sentiment. Under the covers, this app pulls in relevant tweets, calculates sentiment as a percentage of positive words divided by sum of positive and negative words, then visualize the data. Henning pulled data from major cities in swing states and calculated sentiment there as well. But if you were to download the app, you get the results without having to do the data science, because Henning did the hard part for you. Go ahead and check out the results below, but first, a note about sarcasm.
Last night, I posted the image from the Datameer Twitter Election App on Facebook and a friend, Tac Anderson, stated that “I’m almost 100% certain that sarcasm throws this way off”. I’ll let Henning get into more details if he wants, but he used the The Subjectivity Lexicon (list of subjectivity clues) if you want to read more. Come on Tac, Sarcasm wouldn’t play a factor*.
During the Presidential Debate on Oct 16th, overall, Mitt Romney had more positive sentiment especially in Madison, WI and Columbia, SC. Barack Obama had more positive sentiment in Tallahassee, FL, Raleigh, NC and Concord, MA (where Romney was Governor).
During the Alfred E. Smith Dinner, the candidates were funny and fairly nice to each other. The Twitter sentiment analysis shows the same. This was much more positive for Romney than the debate on the 16th.
Last night was much closer in terms of sentiment. Across the board it looks like this will be a very close race.
Without digging into the merits of the actual issues discussed, there are a couple of takeaways the Romney and Obama camps could get from this analysis. If you want people to say nice things about you, be funny and nice. My wife watched the debate on Oct 16th and was disappointed how rude both candidates were to the host. The sentiment on Twitter was much more positive during the dinner than the debates. Perhaps the candidates learned from that and that is why they were more cordial last night. And did you notice their scores were higher last night than from their first debate?
Anyway, analyzing sentiment on Twitter normally takes a Data Scientist and can be difficult without the right tools. Henning did the hard part and made an analytic app with Datameer so you don’t have to start from scratch. And while candidate sentiment is certainly interesting, the best part is every Datameer app is completely open so you can customize the app to understand sentiment of topics you or your company are interested in.
*sarcasm can become a data quality issue