It seems a bit rich for me to be typing something about sarcasm.
Or does it?
That’s just the problem with sarcasm – ultimately, it can just be very difficult to tell whether words are intended to be sarcastic or not, unless you really know they person who said them, and the context in which they were said. And often, who they were aimed at.
Sarcasm has long been a particularly British quirk. We’re pretty good at it, on the whole. People from many other countries are often bamboozled, or even delighted, by it, even if our trait of (perhaps..) never saying what we really mean can be painful for them at times.
It’s a topic that has hit the headlines (no, it really has) again in the past few days after researchers (and I mean that, genuine research people) developed an algorithm that helps to detect intended sarcasm – and other emotional subtext – via the use of emojis.
When sentiment analysis swept into social media management circles in the late 2000s, sarcasm was cited by some observers as an obvious limitation – think of faded brands that try to revitalise themselves, and the challenges they have with conventional sentiment tools in understanding whether mentions on social networks were firmly tongue-in-cheek or not.
So is an algorithm that can sniff out sarcasm a step forward? Well, yes and no. Whatever you might think of emojis, visual communication – and in the case of social media, images that accompany words – can do much to transmit the full meaning. Digital intelligence that can assess intended meaning effectively and weed out incorrect analysis has real value, particularly as the ever-increasing use of emojis has ever-more potential to throw those findings off-course.
The reality though is that – like sarcasm itself – context is all-important. Being able to understand the people doing the communication, and who they were communicating to, is just as important as a deeper understanding of the words and pictures themselves. Meaning that gaining detailed insights into the real human beings uploading those emojis is the bigger prize here.
Analytics is getting there, but there will always be a need for genuine human understanding in parallel. The more niche the topic and the closer the relationships between the people in the conversation, the more likely it is that you have to be in that ‘circle’ to be in on the joke.
We can train computers to understand hidden meaning. Meanwhile, some meaning will always be hidden for computers – and it may just be the more valuable data, that can only be tapped by asking human beings what they actually meant to say.