#Twitter Being Used to Help #Computers Understand #Sarcasm

One of the biggest problems with online, or any kind of text-based communication is that it's very difficult to convey tone. There are workarounds, changes in phrasing, an increase in bluntness and of course, emojis, but some things are still very difficult to convey accurately, and sarcasm is certainly one of them.

Through platforms like Twitter and Reddit, a certain style of writing has developed which behaves like a written cousin to the so-called lowest form of wit, well enough for people to recognise it when they see it, but what about computers? Two researchers from Carnegie Mellon University in Pittsburgh, David Bamman and Noah A. Smith, have just published a paper entitled 'Contextualised Sarcasm Detection on Twitter', sharing the results of a system of computerised sarcasm detection they created, using tweets as a framework.

It shouldn't come as much of a surprise, Twitter has become such a reactionary medium that it's now a hotbed of dry wit and mockery, more so than any other social media platform. Bamman and Smith used a number of different factors, from particular keywords to a seemingly misplaced 'lol' to hyperbole. They also, of course, included #sarcasm, although if you ask me, anyone who drops that hashtag should immediately be discounted, since declaring that you're being sarcastic kind of defeats the entire purpose, just ask Homer.

In terms of keywords, they looked for things like 'dare', 'shocked', 'clearly' and 'gasp' and for hyperbole they made sure that it highlighted intensifiers like 'really', 'so' and 'very'. Beyond phrasing, there were other factors that played into it, like location (sarcasm is more prevalent in some parts of the world than others), as well as age demographic. Presumably anyone anyone who fell with in the generation X age bracket had a Twitter account brimming with sarcasm (yep, two Simpsons clips in one article, you're welcome).

Using these parameters, they managed to get the computer to detect sarcasm with an impressive 85% success rate. The question you might be asking yourself now is - what's the point of all this? Well, primarily, it falls under the Turing test umbrella, enabling computers to understand and replicate human communication, but beyond that, it could potentially make the filtering of social media posts that much more effective, recognising the difference between a harmless joke between friends and legitimate verbal abuse. That's so amazing, oh my god, wow.

Post a comment


Author Name

Free Gift

Free Gift
Get immediate access to our in depth video training on the click by click steps required to get your successful online business started today

Contact form


Email *

Message *

Powered by Blogger.