Geekend: Sarcasm Detector Wanted - InformationWeek
IT Leadership // IT Strategy
07:00 AM
David Wagner
David Wagner
Connect Directly

Geekend: Sarcasm Detector Wanted

US Secret Service wants a bucket for those times you are dripping with sarcasm.

Top 10 Secret Reasons Microsoft CEO Ballmer Retired
Top 10 Secret Reasons Microsoft CEO Ballmer Retired
(Click image for larger view and slideshow.)

Welcome to my first Geekend on I wrote this weekly column for three and a half years on, and now that I'm bringing it to InformationWeek, I'll repeat what I wrote in my very first Geekend:

Every Friday I'm going to talk about stuff I love -- gadgets, games, movies, and the people who make them -- and I'm not going to worry about the enterprise. Just once a week. One post, for a little bit of fun so we can enter the weekend on a high note.

I hope the InformationWeek audience enjoys my little jaunt through the geekier side of life as much as I do.

[How's your company's geek:jock ratio? Read Geeks Versus Jocks: CIOs, Beware Your Culture.]

Speaking of geek life, the US Secret Service is making an interesting jaunt into natural language processing by asking people to make a sarcasm detector. Here's an intercepted video of an early test:

You can see they're still working on it. But seriously, the Secret Service really does want its own social media monitoring software (it uses FEMA's now), and among the criteria -- buried among other humorous requirements, like its needing to be compatible with Internet Explorer 8 -- is the need to detect sarcasm.

Why would anyone need to do that on Twitter? Isn't Twitter the most sincere place on Earth?

As much as I want to make fun of the government's inability to detect humor of any kind, there's a real need here. You know those signs at the airport that say, "All jokes about bombs will be taken seriously"? That's Twitter for the Secret Service. How do they know the difference between a tweet that says, "I want to kill the president over that decision" and "I want to KILL the president over that decision"?

Allow me to share a personal story of this problem before social media even existed. My grandfather was a Navy veteran and literally would have taken a bullet for any US president, regardless of party. But he also considered it important to write letters to express his dissent. Once he wrote a letter to President Reagan (whom he loved) protesting something Reagan was doing with veterans' benefits. He wrote: "A man ought to be shot for thinking like that..."

Uh oh. Somewhere in Washington, a little file was created. And several years later, when Reagan was visiting my grandfather's hometown, he got a knock on the door. The Secret Service agents told my grandmother they'd be outside the house until the president left. My grandfather was confined to his chair because of a stroke, and when my mother informed the agents, they went in, politely talked with my grandfather, and left.

But imagine this encounter on the scale of Twitter. Not just a few thousand angry letters, but 500 million tweets per day, plus Facebook, Tumblr, Instagram, and countless other social forums. And they aren't all hashtagged #Iwanttokillthepresident #thisisnojoke. How do you tell the difference? How do you even read them all?

So good for the Secret Service, but this is a tall order. How do you teach a computer something that not all people are good at?

We're getting surprisingly close (and yet still so far away), with teaching computers about regular humor. We've even got them writing jokes like this one: "What do you get when you combine a fragrance with an actor? A smell Gibson." That at least resembles a joke. But how do you teach sarcasm, which requires an understanding of the intent

Next Page

David has been writing on business and technology for over 10 years and was most recently Managing Editor at Before that he was an Assistant Editor at MIT Sloan Management Review, where he covered a wide range of business topics including IT, ... View Full Bio
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.
1 of 2
Comment  | 
Print  | 
More Insights
Newest First  |  Oldest First  |  Threaded View
<<   <   Page 7 / 7
User Rank: Ninja
6/6/2014 | 10:45:19 AM
Re: Sadly we do need this
Haha! Curious view of natural language processing :) Yes, it does seem as though plenty of government folks are reading "great job guys!" has true and honest not as sarcasm.
User Rank: Ninja
6/6/2014 | 10:04:09 AM
Re: Sadly we do need this
@dave – Bazinga! Sentiment detectors like Lymbix and others comb through communications for sentiment and tone.  Of all the emotions,  detecting  sarcasm is particularly difficult, since it requires a great deal of context about the writer's subject, as well as an understanding of the use of language and paralanguage (images, etc). But, it can be done, and has been the subject of considerable work in computer labs since the 1970s.

It buys you the same thing that understanding any other emotional state. Emoticons are among the many ways people indicate sarcasm. Watch out for those winking ;-) faces . So, for the foreseeable future, the more textual communication we have, in social media and email, the more we need to understand what it all means in a context that makes sense. Challenging.

So, maybe not so sadly, we do need this.
User Rank: Ninja
6/6/2014 | 7:34:00 AM
Sadly we do need this
My first thought was wondering how often our government hears "great job on that one guys" and takes it as a compliment rather than sarcasm.  Now I'm pretty sure that some politicians will always spin sarcasm to be a compliment but government agencies really need to be careful about how they take what is said/written/tweeted.  After they get the sarcasm detector figured out the next thing they need to work on is a BS detector so that they can stop taking people who are just spouting off so seriously.  I see a lot of dumb things repeated and wish there was a good way to mark garbage tweets as what they are.
<<   <   Page 7 / 7
Register for InformationWeek Newsletters
White Papers
Current Issue
Digital Transformation Myths & Truths
Transformation is on every IT organization's to-do list, but effectively transforming IT means a major shift in technology as well as business models and culture. In this IT Trend Report, we examine some of the misconceptions of digital transformation and look at steps you can take to succeed technically and culturally.
Twitter Feed
Sponsored Live Streaming Video
Everything You've Been Told About Mobility Is Wrong
Attend this video symposium with Sean Wisdom, Global Director of Mobility Solutions, and learn about how you can harness powerful new products to mobilize your business potential.
Flash Poll