On Saturday, we released an infographic about what we call the “Social Oscars” with TechCrunch. We looked at social media mentions around the movies nominated for Best Picture. We measured the volume, built a sentiment index and more interestingly, detected the number and type of unique adjectives used in those social media posts. For The Artist, we found adjectives like “unprecedented”, “predictable” or “perfect film”. We were not trying to predict the actual Oscar winner, but to understand how movie-goers felt about each movie. Obviously, movie fans don’t get to vote in the Oscars.
After looking at the number of adjectives and their type (positive/negative), the number of mentions as well as the sentiment index, we decided to award the “Social Oscars” to The Help. The Artist had more mentions than The Help (69,000 to 52,000) and a higher number of unique adjectives (162 to 68). But The Help had a higher sentiment index by far (92% to 79%) and a higher percentage of positive adjectives.
But this opens up a very interesting issue. Can you use social media data to predict the results of certain events? And if so, what types of events? Can you use it to predict the Oscar winners?
It turns out, there’s quite a bit of research being done in this field. Researchers have studied whether tweets can predict which scientific papers are going to be cited the most (“highly tweeted articles were 11 times more likely to be highly cited than less-tweeted articles“). Tweets also seem to correlate to how the stock market is going to perform (“We find an accuracy of 87.6% in predicting the daily up and down changes in the closing values of the DJIA and a reduction of the Mean Average Percentage Error by more than 6%.“). Looks like you can use tweets to predict the box office success of a Hollywood movie (“showed that the results outperformed in accuracy those of the Hollywood Stock Exchange“) but not the result of elections (“data from social media didonly slightly better than chance in predicting election resultsin the last US Congressional elections“).
So what do all these have in common?
Regardless of the field being researched, social media was able to predict the outcome of an event when the type of people doing the posting also had a means of influencing the outcome in real life. If I tweet about a movie before its launch, that means I might go see it, thus influencing the box office results. This opens up amazing possibilities. You can potentially predict how a new product is going to do in a new market. Or what book is going to become a best-seller. Or which TV ad is going to bring in the most sales.
So then, what about events like the Oscars or the Superbowl? As long as regular people don’t get to vote in the Oscars or play in the Superbowl, they don’t really have a way of influencing the result. So using social media to try to predict that is like using tweets to predict a roll of the dice. You might get it right every once in a while, but scientifically, it’s not better than chance. All one can say then is, if people got to vote, who might be the winner of the “Social Oscars”? Which is what we did.