A post on Popsci tells of how an Israeli team have created an algorithm that will allow computers to detect sarcasm amongst the wider patterns of normal speech.
SASI, a Semi-supervised Algorithm for Sarcasm Identification, can recognize sarcastic sentences in product reviews online with pretty astounding 77 percent precision. To create such an algorithm, the team scanned 66,000 Amazon.com product reviews, with three different human annotators tagging sentences for sarcasm. The team then identified certain sarcastic patterns that emerged in the reviews and created a classification algorithm that puts each statement into a sarcastic class.
The recognition of language formation is one of the key barriers to fully automated monitoring tools, especially in the social media space. You always need a human to sense check any sort of content and sentiment analysis.
Despite the flippant comments below the post, this is a hugely fascinating development. It is easy enough to associate the word ‘great’ with something positive, it is also sensible to associate ‘bad’ with something negative.
But, what if you switch the established meanings of those two words around?
The sentence “The storm was just great, I got caught in it.” could be read both positively and negatively. Likewise with “I went to see that film, it was well bad“.
It’s the context around the phrase that defines it’s meaning and use in the sentence. This is something that so far, our friends ‘the programmers’ haven’t yet quite mastered.
If a technology can get within 10% of the overall sentiment of a set of linguistic data, that’s cause for alarm for those in the online monitoring space.
Here’s what they might think:
1) Blimey, we might be out of a job
2) How can we better that?
3) How much is it and can we integrate it?
How much time is spent collating social media data and assessing tone and sentiment in digital agencies? Probably a lot.
If you’ve got a computer to do that for you, AND it’s reliable, it’s another nail in the coffin for the traditional AE job of monitoring and reporting, something I think many would look forward to.
