Customers post good and bad opinions of products and services – and they also share witty and sarcastic opinions. All of which have been notoriously difficult for brands to measure. Sentiment analysis in the social listening world has not enjoyed the best of reputations as a lack of any advanced technology has meant that accurate results have depended on a process of repetitive, tedious and time-consuming manual tagging of documents. As a result, many brands tend to sideline sentiment indicators in their reporting. Now all that is set to change, says Richard Sunley. Machines have been learning how to fathom irony . . .