Languages are fraught with nuances, but the one language technique that gives people problems understanding – let alone computers – is sarcasm. Despite this obvious degree of difficulty, researchers at the Carnegie Mellon University are developing a system that can identify sarcasm on Twitter. As detailed in their paper titled “Contextualized Sarcasm Detection on Twitter”, authors David Bamman and Noah Smith break sarcasm down to its roots, identifying that sarcasm can be better identified if we can understand the context in which it is used in. For instance, the use of the hashtag, #sarcasm, on Twitter is not necessarily an indicator of the use of sarcasm, but a signal to those reading the tweet that what they are reading is the true meaning of the tweet, rather than the implied meaning as is usually the case with sarcasm. Confused? Me too.
Even though this seems like a pretty impossible task for a computer to do, the algorithm used by Bamman and Smith has managed to achieve 75% accuracy when it comes to identifying sarcasm if considering only the tweet. This accuracy rises to 85% if the algorithm is allowed to also consider who the author is, their audience, and the response to the tweet. Obviously it’s not a perfect system yet, but it’s well on its way to being useful. You may be asking why this is even important: the simple answer is this could be extremely useful in determining whether threats made on social media, like Twitter, are genuine threats or simple sarcasm.
What do you think about this research about sarcasm on Twitter? Let us know your thoughts in the comments below.
