Researchers at Massachusetts Institute of Technology (MIT), USA, have developed an algorithm which can detect sarcasm in tweets, apparently better than most people.
The researchers had initially aimed to develop an algorithm that can detect racist and abusive content but in the process developed this algorithm first as they felt it’s important for the machine to understand sarcasm.
Researchers believe that understanding of sarcasm is the first step for the algorithm towards gaining a better grasp of the emotional subtext of a sentence.
Also on Guiding Tech
“Because we can’t use intonation in our voice or body language to contextualize what we are saying, emoji are the way we do it online,” says Iyad Rahwan, an associate professor the MIT Media lab who developed the algorithm with one of his students, Bjarke Felbo, told MIT Review.
“The neural network learned the connection between a certain kind of language and an emoji,” Rahwan added.
Twitter is already a hub for trolls and the company has been ramping up its efforts to curb the menace.
Gauging attitudes and behaviour of people towards posts on social-media has been a prevalent practice among advertisers.
When fully developed, this algorithm can prove to be elementary in helping quash abusive/racist/harassing tweets and the users too.
The algorithm uses deep learning technique which trains a simulated neural network to identify and understand patterns using large amounts of data.
The researchers used a very common way of showing emotions on the internet — emojis — as a labelling system and one of the ways of training their algorithm to identify emotions in tweets.
To test out the bots in the real world scenario against humans, the researchers recruited volunteers through crowdsourcing website Mechanical Turks. The algorithm identified sarcastic undertones in tweets with 82 percent accuracy as compared to the human volunteers who identified sarcasm with 76 percent accuracy.
“It might be that it’s learning all the different slang,” Felbo says. “People have very interesting uses of language [on Twitter]—let’s put it that way.”
The researchers collected over 55 billion tweets in all, with 1.2 billion of them containing emojis. Using these emoji embedded tweets, the researchers helped the algorithm learn and identify that which emojis are used with which kind of text — happy, sad, humorous and so on.
Computers are getting better at machine learning day by day and are getting a better sense of how humans talk and behave via social media data mining.
This algorithm can be used to curb abusive, racist and terrorism-related content from not only Twitter but the other organisations such as Facebook, YouTube, Snap and others which are trying to make their platforms as well as the internet a better place.