Sarcasm when done in real-life can be pretty obvious as it is often accompanied by eye-rolling, the tone of voice, and body language to indicate that what was said wasn’t said in seriousness. Unfortunately sarcasm online doesn’t translate too well. In fact a couple of years ago, the US Secret Service wanted sarcasm detectors on Twitter.
Turns out that they might be getting their wish, thanks to researchers who are trying to teach computers how to detect sarcasm on social media, like Twitter. The paper is titled “Contextualized Sarcasm Detection on Twitter” and basically it trains computers to look out for keywords that might denote a sarcastic message.
Such words include “clearly”, “shocked”, “gasp”, and “really”. It also pays attention to hashtags like “#lol” and the dead giveaway, “#sarcasm”. The computer was also trained to pay attention to other factors like location, age, gender, history of posts, and more. According to the researchers, it seems that they’ve had some success with it and have reported an 85% success rate when it comes to detecting sarcasm online.The paper doesn’t dive into what kind of practical applications one might be able to expect from such technology, but like we said before, this could be used by law enforcement like the US Secret Service to determine which threats are real, and whether a post was made out of frustration or a joke in poor taste.