Tumblr 101: Tips And Tricks To Get Started
Facebook Plunges Again
Computerized detection of sarcasm is closer than you think
"In many cases, sarcasm is difficult even for people to recognise," says Ari Rappoport of the Hebrew University of Jerusalem, Israel. No shit, Ari, really?Given a sample set of Amazon product reviews and random Twitter posts, the new sarcasm software agreed with human analysis more than 75% of the time!
While this is a pretty nifty advance for natural language processing, the implications are less obvious. Today, computerized sarcasm detection is only really useful for the tracking of online brand sentiment by marketers. Tomorrow though... maybe it would be useful to have a robot assistant that knows when you're being sarcastic? 'No, I need your help to wipe my own ass.'
Or perhaps, if a computer can detect sarcasm, it could also remove it? Imagine having such software on your phone -- scathing sarcasm to your significant other would become a thing of the past! 'Yes, of course I'll just drop this deadline and and be home for dinner sweetie!'
When it comes down to it though, this is all just a frivolous exercise in exorbitant experimentation -- the SarcMark is a far more efficacious and judicious use of our hard-earnt money.
Comments
9
Subscribe to commentshazardMay 26th 2010 9:34AM
I think a lot [not all] of North Americans need the wetware version this as 75% is a lot better than 0
Sebastian AnthonyMay 26th 2010 9:48AM
See... I wish I was allowed to make comments like that :(
MJGMay 26th 2010 10:38AM
Oh, a sarcasm detector, that's a real useful invention.
Sebastian AnthonyMay 26th 2010 8:38PM
I see what you did there.
put-itMay 26th 2010 1:39PM
Because we have no other problems
fincanMay 26th 2010 2:33PM
one the funniest ds articles in a long time.
Sebastian AnthonyMay 26th 2010 8:38PM
See, I know you're not being sarcastic because it's TRUE!
SalmanMay 27th 2010 12:04PM
SUCH a useful application...
GeorgeMay 27th 2010 4:42PM
I think the wording is much more exact in this post's title than in the assertion "In many cases, sarcasm is difficult even for people to recognise", because nobody has come even close to having the computer approximate in any meaningful sense what humans are doing when they interpret (or fail to interpret) sarcasm. Mixing both things in the same statement seems extremely misleading. Basically, what the authors of this research have achieved is the same as using email filtering technology: humans train the system giving it the discrimination criteria (for instance, an email address), and the system performs the discrimination accordingly. I am afraid there may be no single line of actual science to it. Without a case study, there is no way to tell, it may well be that those 77% of sarcastic messages contained a strong cue, for example a characteristic emoticon or an expression such as "Great" without an exclamation, and that strong correlation may account for most of the figure, with no additional insights of interest (it would merely be a case of emoticon matching technology having being invented, which amounts to discovering and automating the "Find" feature in any word processor).