An Isreali research team has developed a machine algorithm that can recognise sarcasm. In tests, the Semi-supervised Algorithm for Sarcasm Identification (SASI) has identified sarcastic sentences in online product reviews to some 77% accuracy.
The team developed the algorithm by scanning 66,000 Amazon.com reviews with three humans tagging sentences for sarcasm and putting each statement into a sarcastic class. The algorithms were trained on a test set of sentences and “learned” the sort of words and patterns in sarcastic remarks – i.e. remarks whose literal meaning is opposite to the sentiment conveyed.
Programs that can recognise sarcasm could be used to generate personalized content, differentiating between statements such as “S Club 7’s Greatest Hits is the most important record of all time” and more constructive reviews.
* * *