2 min read

Xerox is building a sarcasm detector? That’s a great idea


If there’s one thing citizens of the modern age know, it’s hard to detect tone in an e-mail. If my boss really mad at me, or am I reading too much into this message? Well, if it’s hard for a human to detect tone, imagine the difficulty for a machine.

It’s an issue to be addressed as automation increasingly comes to areas such as social media monitoring and analytics, driven by the increasing use of big data, as marketers try to harness the conversations happening across Facebook, Twitter, YouTube, LinkedIn, blogs and other online communities. Is a customer tweeting about their experience happy? Are they frustrated? And what should be done with that information to improve the customer experience?

It’s something researchers at Xerox have been working on, including developing an automated data analytics platform that can more accurately determine the sentiment of comments, and automatically route it to the right person for action. If a customer is angry, for example, you want to get to them quickly and try to resolve the issue.

“When humans have to step in and evaluate the context of a tweet or route a post, it slows things down and reduces the overall value of social media data,” said Tong Sun, head of the data analytics laboratory at Xerox Research Center Webster, in a statement. “We’re piloting a platform that lets computers do the heavy lifting.”

The idea is to use automation to help a company address an issue before it becomes a crisis. Xerox says a recent pilot saw business response time cut to hours from days or weeks.

“Studies have shown responding quickly to an issue raised on social media channels leads to happier customers,” Tong said. “And we know that targeting the right customers and communities makes for a more profitable marketing campaign.”

Rather than using a keyword-centric approach to determine sentiment like many social media monitoring tools, which can be thwarted by sarcasm and abbreviations, Xerox researchers with with expertise in text mining, machine learning and predictive modeling built a new analytical platform that seeks to determine sentiment within the context of ongoing conversations, assesses the priority, and sends it to the right contact for action.

Xeros said the technology is now available as part of its Xerox Customer Care offerings, and is being targeted at organizations including the financial services, telecommunications and retail industries.