Oh, great. Now the one thing that makes humans such unbearable twats will be lost forever to a bunch of nerds with math skills.
These scientists were able to use the algorithm to detect sarcasm in 77 percent of 66,000 reviews on Amazon. That's pretty damn good. They did this by setting the thing loose on reviews and trained it to detect sarcasm in reviews marked by humans.
But their academic paper's title? It's called "A Great Catchy Name: Semi-Supervised Recognition of Sarcastic Sentences in Online Product Reviews."

Oh joy. This is a really important step forward for Science.

77% sarcasm detection accuracy rate? That's better than most of the people on this forum!

Haha.. That was really funny!!!!!!!

Meh, not impressed. It's basically an algorithm that looks for certain phrases & punctuation. I think if they had tried this on the British Amazon, the results would have been a lot lower. Americans tend to make REALLY obvious sarcastic comments using exclamation marks, classic sentence structures (like "Gee, I sure like ... ") and, just in case you didn't get it was sarcasm, mention it just in case (*sarcastic voice*). Brits use a dryer form of sarcasm that is a lot harder to detect.

Agreed.
It would have been more impressive if it managed the detection by a semantic analysis. This basically is a relatively simple pattern recognizer that would only work in languages where sarcasm is as explicitly marked as in American Internet English :p

I DETECT SARCASM

But you can't be 77% sure unless you use that new algorithm!

They should also test the algorithm on their paper's title.

If you'd try this algorithm on a bunch of average "web 2.0"-"in"-blogger-blogs about political subjects, it could get a 95% success-rate just by marking them all as sarcastic... :P

Aided by Dr. Sheldon Cooper, probably. A Cooper / Frink project, of course, to keep things in alphabetical order.

I think drier sarcasm is easier in some societies. I used to be very dry in my sarcasm until I got my current job; too many people thought I seriously meant the things I said.
I do question whether this can be done successfully by software. In the place and time I grew up, I was taught that "sir" and "ma'am" were how one addresses people with respect and in formal situations like doing business or with someone you don't know well. In other parts of the country, people get offended because it almost always used sarcastically. How is a program going to account for place and time like that?

True, a lot of sarcasm is context dependent. Proper interpretation of such would probably require a full blown AI. And how can any algorithm go up against Poe's Law?

You are all so naive and simple minded. You just can't get the genius of this advancement and what it will mean for the human race. Just wait till a model T comes breaking through your door and asks if you want some lead with your coffee.