You Don’t Say? Computers, Science and Sarcasm

by Christian Gibbons
Summer 2016 Intern
 
Like many other people, I have experienced times when a sarcastic comment has flown right over my head. The thing about these moments, though, is that when you don’t pick up on sarcasm, you tend to miss a lot. Who knew?

A big reason why it’s so important to be able to recognize sarcasm is because of how common it is in our society today. Over the course of the development of the English language, entire phrases that were once meant sincerely have come to have an exclusively sarcastic meaning. After all, when was the last time you heard someone say, “Big deal!” and really mean it?

In fact, sarcasm is so omnipresent in our society that computers need to be able to recognize it, too. A couple of years ago, the US Secret Service began seeking a software system that would be able to distinguish sarcasm. One of the Secret Service’s spokesmen, Ed Donovan, stated in 2014 that the agency’s ultimate goal was to analyze social media data, and that “the ability to detect sarcasm and false positives is just 1 of 16 or 18 things [they were] looking at.”

There are clear and immediate benefits to a tool like this. Perhaps most importantly, a sarcasm detector would allow people to differentiate between an instance of sarcasm and a legitimate threat on social media platforms such as Twitter. The problem is, computers that are used as sarcasm detectors often have a difficult time identifying such a nuanced form of communication.

However, researchers at Carnegie Mellon University recently published findings that might be on to something that will change that. What made this study different from others was that its authors believe sarcasm is more likely to be used between people who know each other well. Using algorithms that took into account contextual features such as past tweets, key words and phrases, profile information, and previous sentiments, the researchers found that their detectors were able to pick out sarcasm with increasing accuracy over time. The highest accuracy rating they achieved overall was 85.1 percent!

As scientists continue to work on the challenges of getting computers to understand sarcasm, who knows what level of accuracy they’ll be able to achieve? Maybe some day computers will be able to recognize sarcasm faster than humans can. That wouldn’t be creepy at all.

Did You Know?

Sarcasm may have a sour reputation in some circles, but scientists have found that there are actually benefits to using it. Because understanding sarcasm involves disparities between literal and intended meanings, sarcasm seems to promote greater creativity in both people who use it and people who hear it.