We take sarcasm quite naturally and usually don’t need any help figuring it out, like, in a social media post. Machines, on the other hand, will have a hard time on this as they are only programmed to read a text and evaluate images closely based on what they see. But what’s in it for us? Nothing, except if some great scientist will help computers improve its brainpower to understand our overly ironic wordplay in social media and in the internet as a whole. But it appears they’re about to make it real.
Thanks! That’s Just What I Need!
Yeah, right. It’s just what you need –a “sarcasm-detecting” software that can aid marketers in determining whether you’re praising or insulting their products and from it, adjust their messages to peddle you with more stuff. However, promoters add that these savvier computers may also help law enforcement agencies, as well as IT Security Services Australia, characterise real threats from those that embellish or ridicule at serious topics, especially on social media sites like Twitter, Instagram, and Tumblr image posts. It might even help ACS systems to figure out if their customers are upset, and direct you to a real customer support rep from IT Services Australia or allow politicians to gain insight whether their messages reverberate to the voters.
A computer science professor at the University of Turin and a group of colleagues from Yahoo! are attempting to teach machines that humans doesn’t always mean what they say. What’s new about their research is that they utilised images and texts to look for clues and understand its meaning. They said that looking at a text isn’t enough and that the images can provide more leads.
This Must Be Really Important For You!
To deal with the problem of converting this kind of nuance into something that machines can understand, the research team ran to humans for help. With the help of Yahoo! researchers, they designed a crowdsourcing tool to ask English-speaking people from different countries to tag any social media posts they see as sarcastic or not. First, they analysed text-only posts, then statements accompanied with images. The researchers found out that, despite participants not agreeing which posts are sarcastic or not, statements accompanied by visual images helped identify a double-edged message. Regardless if there is an accompanied image, linguistic cues that signal sarcasm includes wordplay and punctuations, particularly using exclamation marks.
Then the researchers wrote a computer algorithm that mathematically represents the humans that taught them. This will allow the machine to use the provided data to check on posts for sarcasm. By utilising a combination of features, the machine was able to achieve 80 to 89 percent rate of catching sarcasm on every post it scanned. But the accuracy drops when used on posts with visual forms or images.
Sarcasm Is A Lower Form of Wit
Enhanced machine processing power and extensive social networks can make this type of machine learning possible. More powerful processors can effectively handle this kind of learning, with social networks providing the data.
Some scientists in this field conclude that the said research is an important leap in literature because studies on machines ability in identifying natural language as sarcasm or irony surely require advanced methods and processing.