Slashdot Mirror
Software Recognizes Sarcastic Tweets
An anonymous reader writes "Even humans sometimes fail to recognize sarcasm and irony; can machines do better? An algorithm that identifies sarcastic tweets (PDF) on Twitter and sarcastic sentences in product reviews on Amazon will be presented next week in the International Conference for Weblogs and Social Media in Washington, DC, and in the Computational Natural Language Learning in Sweden in July. A team from the Hebrew University, Israel, has developed an algorithm that identifies sarcastic sentences by using a machine learning technique in which a small number of sarcastic sentences act as seeds for the software to learn and generalize upon. The algorithm can then identify sarcastic sentences that are nothing like the examples. The variety of recognized sarcastic sentences is impressive, though the results are not perfect. But again, we don't do it so well ourselves, do we?"
Yeah, sure it does.
Want to improve your Karma? Instead of "Post Anonymously", try the "Post Humously" option.
Weight of various patterns and features. We present here a deeper look on some examples. A classic example of a sarcastic comment is: "Silly me, the Kindle and the Sony eBook can’t read these protected formats. Great!". Some of the patterns it contains are ...
You know DRM is pervasive as a very serious consumer problem when statistical research papers recognize user dissatisfaction with it as a classic example of sarcasm that floods reviews.
My work here is dung.
I love links to PDFs.
Oh, that's really useful!
This may help people with autism and Asperger's Syndrome recognize satire.
homo ergaster,
I am your master !
What great grammar skills you have.
A sarcasm detector, that's a real useful invention!
I guess it's for those who just don't get it!
"Yeah, we totally developed a program to detect sarcastic tweets... #fuckinggenius"
Just end your sarcasm tags before being sarcastic. This won't conform to W3C standards, however.
"Hegelians, who love a synthesis, will probably conclude that he wears a wig." - Bertrand Russell
Given that sometimes not even humans understand when I am being sarcastic, I expect this software will have an exceptionally high recognition rate with very low false positives. A truly remarkable achievement and the one algorithm the human race has been waiting for!
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
The algorithm can then identify sarcastic sentences that are nothing like the examples.
Good luck with that.
I can see the fnords!
wow
I'm god, but it's a bit of a drag really...
I forsee nothing but success for this algorithm.
"Any sufficiently optimistic statement is indistinguishable from sarcasm."
A sarcasm detector, that's a real useful invention!
It's only a matter of time before we can automatically generate sarcasm. Then websites can have snide comments auto-generated. When that happens, I'd like to see penalties for those attempting clever snark but failing to be smarter than a computer.
What about someone whose intent is genuine, but who can only speak with a sarcastic tone of voice?
http://www.youtube.com/watch?v=PyujQctZ9hw
Yeah, like that would work!
-- Two men say they're Jesus. One of them must be wrong. - Dire Straits
The algorithm can then identify sarcastic sentences that are nothing like the examples.
Place it in my office. If it still responds at the end of the week, it's not working correctly. If it's overloaded and partially melted, we've got a winner.
They got about 80% accuracy. Looks like things in CAPS and other... types! of punctuation ;-) and metadata (star rating given for review) are used fairly heavily in addition to sentence structure. Would have be good to know what the breakdown of false positive vs. missed sarcasm is, but i didn't see it.
From TFA:
"We experimented with a large data set of 66000 reviews for various books and products. Evaluating pattern acquisition efciency, we achieved 81% in a 5-fold cross validation on the annotated seed, proving the consistency of the pattern acquisition phase. [...] each sentence was annotated by three human readers. We found some strong features that recognize sarcastic utterances, however, a combination of more subtle features served best in recognizing the various facets of sarcasm."
You won't hear about this in the media, but I still wish briefly to take a position on the question as to what extent Newt Gingrich's dodgy smears induce paralysis of the cerebrum. Read on, gentle reader, and hear what I have to say. When a friend wants to drive inebriated, you try to stop him. Well, Gingrich is drunk with power, which is why we must open students' eyes, minds, hearts, and souls to the world around them.
Gingrich has a glib proficiency with words and very sensitive nostrils. He can smell money in your pocket from a block away. Once that delicious aroma reaches Gingrich's nostrils, he'll start talking about the joy of irreligionism and how people are pawns to be used and manipulated. As you listen to Gingrich's sing-song, chances are you won't even notice his hand as it goes into your pocket. Only later, after you realize you've been robbed, will you truly understand that this is partly connected with what I wrote earlier concerning soporific dirtbags. Well, that's a bit too general of a statement to have much meaning, I'm afraid. So let me instead explain my point as follows: He likes to posture as a guardian of virtue and manners. However, when it comes right down to it, what Gingrich is pushing is both grungy and antihumanist.
I receive a great deal of correspondence from people all over the world. One of the things that impresses me about all of it is the massive number of people who realize that Gingrich has written volumes about how his strictures provide a liberating insight into life, the universe, and everything. Don't believe a word of it, though. The truth is that I do not propose a supernatural solution to the problems we're having with him. Instead, I propose a practical, realistic, down-to-earth approach that requires only that I focus on concrete facts, on hard news, on analyzing and interpreting what's happening in the world. I've never bothered Gingrich. Yet Gingrich wants to foster and intensify his drug-drenched drama of immorality. Whatever happened to "live and let live"?
I've long thought it would be fun to try to explain to Gingrich how he is just making a mug of himself when he says that officious Luddites have dramatically lower incidences of cancer, heart attacks, heart disease, and many other illnesses than the rest of us. For the most part, I'm just curious as to how deep Gingrich will have to dig into his profanity thesaurus to formulate a response. He seizes every opportunity to lead people towards iniquity and sin. I cannot believe this colossal clownishness. Any sane person knows that there is a problem here. A very large, amateurish, disreputable problem.
What I think—and I'm no specialist—is that I am deliberately using colorful language in this letter. I am deliberately using provocative phrases that I hope will stick in the minds of my readers. I do ensure, however, that my words are always appropriate and accurate and clearly explain how this is not Nazi Germany or Soviet Russia, where the state would be eager to dismantle the guard rails that protect society from the superficial elements in its midst. Not yet, at least. But it doesn't do us much good to become angry and wave our arms and shout about the evils of Gingrich's suggestions in general terms. If we want other people to agree with us and join forces with us, then we must debunk the nonsense spouted by Gingrich's secret police. I won't lie to you; I, not being one of the many barbaric election-year also-rans of this world, don't care what others say about Gingrich. He's still infernal, paltry, and he intends to quote me out of context.
It's not easy for me to say this, but I wouldn't put it past Gingrich to reinforce the concept of collective guilt that is the root of all prejudice. There, I said it. Now I can continue with my previous point, which is that I recently heard a famous celebrity—I forgot which one—say, "Gingrich's 'leave behind a legacy of perpetual indebtedness in developing countries' mentality is so pervasive that I feel like I'm going to
Recognizing sarcastic tweets... do twits even know what sarcasm is?
Slashdot social media options: AIM, ICQ, Yahoo, Jabber and Mobile Text. Why no MySpace?
I wonder what would happen if you applied their algorithm to any given slashdot post, particularly one on the Apple board.
-- Two men say they're Jesus. One of them must be wrong. - Dire Straits
What is the name of that phenomenon where it becomes impossible to distinguish actual right-wing views from sarcastic parodies?
We should just do away with paper money and only use gold.
The United States is a Christian nation, so we should teach in public schools that the Earth is only 6,000 years old.
Our president is likely an Islamic terrorist sleeper agent planted here decades ago in order to become president and enact terror-friendly laws.
Step into a huge movement. Don't Tread In Me.
This really *is* useful. (Something for the detector: I read TFA but it's just because I can't resist the elegant typesetting of PDFs.)
In reality though, automated systems that process a large amount of social comments (think Amazon reviews etc.) can be fooled by sarcastic comments. Such a system could result in poor recommendations.
This is not intended to be useful to humor impaired individuals.
When you hear hoofbeats, think horses, not zebras
A sarcasm detector, that's a real useful invention. (Sarcasm detector explodes)
They should just train it on Slashdot comments.
Many Slashdot users can benefit from such a technology. There are medical conditions whose sufferers cannot detect sarcasm. This leads to social ostracism and can cause reduced productivity and in very extreme cases, depression. People may scoff, but imagine if we could provide this technology on a portable device for those victims of the spectrum of diseases that cause anxieties? There are times when I have been the goat because I misunderstood "Yes, I'll have the work completed by Monday" to mean that the coming Monday, the work would be completed. My sarcasm detector did not fire to alert me that the cable installer was being sarcastic. I would like to see this sarcasm detector available for handheld devices. When a girl responded, "Yeah, I'll go out with you," I could then check my iPhone or Droid and know immediately she was making an attempt at humor.
Found the other stats after taking a second look at the PDF. was 11% false positive, 12% false negative.
Ye right!
__
Sigs are like arse-holes, everybody has one
It dates back into some of the great classic works of our time... upon reading Romeo and Juliet one critic was overheard saying:
"Nice play Shakespeare..."
or upon solving a great mystery, Watson was once overheard saying, "No shit Sherlock."
That condition that they have is actually called mental retardation.
Mit der Dummheit kämpfen Götter selbst vergebens
Thank you! It's Poe's Law.
Step into a huge movement. Don't Tread In Me.
Vercotti: Well, I had been running a successful escort agency - high class, no really, high class girls - we didn't have any of that. That was right out. And I decided. (phone rings on desk) Excuse me (he answers it) Hello......no, not now......shtoom...shtoom....right......yes, we'll have the watch ready for you at midnight.......the watch.....the Chinese watch....yes, right-oh, bye-bye mother (he replaces reciever) Anyway I decided then to open a high-class night club for the gentry at Biggleswade with International cuisine, cooking, top-line acts, and not a cheap clip joint for picking up tarts, that was right out, I deny that completely, and one night Dinsdale walked in with a couple of big lads, one of whom was carrying a tactical nuclear missile. They said I'd bought one of their fruit machines and would I pay for it.
Interviewer: How much did they want?
Vercotti: Three quarters of a million pounds. Then they went out.
Interviewer: Why didn't you call the police?
Vercotti: Well I had noticed that the lad with the thermo-nuclear device was the Chief Constable for the area. Anyway a week later they came back, said that the cheque had bounced and that I had to see Doug.
Interviewer: Doug?
Vercotti: Doug (takes a drink) I was terrified of him. Everyone was terrified of Doug. I've seen grown men pull their own heads off rather than see Doug. Even Dinsdale was frightened of Doug.
Interviewer: What did he do?
Vercotti: He used sarcasm. He knew all the tricks, dramatic irony, metaphor, bathos, puns, parody, litotes and satire.
....
No folly is more costly than the folly of intolerant idealism. - Winston Churchill
I bet it just looks for itallic text.
No one would post a sarcastic remark
Would this help Paul Chambers, the man who was found guilty of sending a menacing messages for his sarcastic Twitter bomb threat? http://www.facebook.com/group.php?gid=265300406002
"I thought they were the dominant species..."
... why can't you?
[signature]
Use the word 'fuck' in your tweet/post. This way the profanity filter will block your message/post before it ever gets to the irony filter..
this scene
With verbal communication, you can indicate sarcasm by tone of voice.
With written communication, you have the conundrum that nobody would understand that you are sarcastic when you write an opinion unless you drop written clues. And there's only so many clues that convention can take for sarcasm. e.g.
'The press release made me simply explode with excitement'. -uneven pattern of peaks and valleys
'This is excellent!!11!1oneone' -the same
'John has said he regrets what he did. Of course.' -double emphasis creates the question of the opposite
Put the set of clues into a machine and you've got your sarcasm detector.
Sarcasm? On *Twitter*?! Never!!!
org.slashdot.post.SignatureNotFoundException: ewg
Their algorithm gets 77% accuracy. I think I can do better:
# Estimated accuracy: 92.1%
isSarcastic(tweet) { return true; }
Or does that only work for slashdot comments?
It will work perfectly.
The point of Sarcasm is that the words, the text itself, convey a literal meaning, while the actual intent (which must be deduced by the reader knowing certain things about the writer; sometimes just tone of voice is enough) is the polar opposite. Without anything except one line of text, there is absolutely no way of determining whether something is sarcasm or not. It will never work without more input. Now that you've read this paragraph, re-read the sentence above it.
@"Like the fair, independent, open-minded, thoughtful people that you know they are."
You think thats bad. With this automated way to effectively profile comments, sites will be able to bias, push down out of sight or even totally suppress comments they don't want on their products.
Then there's the evolution of this technology, which is to profile sarcastic political comments on known political topics to workout people's political affiliations, so the people in power can hold back opponents and help supporters of their party. (Thats the way police states have grown powerful for a long time, but this kind of automated profiling is another step to a level of profiling beyond anything the world has ever suffered before).
Evolution of this research could even form the basis of a system of Thought Crime detection.
(By the way, the sampling period is the key to Big Brother monitoring. (This is why George Orwell's 1984 book showed the power of monitoring someone over their entire lifetime. It showed that once someone was minored for long enough, they could be exploited, controlled and manipulated by someone using their fears and desires against them). So you can obfuscate a few communications and in the process maybe make it confusing for (human) readers over days, weeks even months, but you cannot keep up totally random comments for years and even decades. So over longer sampling periods clearer signals will emerge from your apparent short term chaos. (It simply becomes a clearer profile signal over a larger sampling period)).
. . . that the people who invented sarcasm (i.e.:"Moshe, were there not enough graves in Egypt that you brought us to the desert to die?" Shemos 14:11) would be the ones who find a way to automatically identify it.
In other news, there is still no software that can detect tweets without sarcasm.
Evil is as eval("does");
Its got a pretty web gui and everything.
http://www.twitter.com/
Yes, I'm trolling, but its still true!
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
Please don't sue me twitter, it was a joke, I'm really not claiming I wrote your crappy software!
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
Irony is easy. Ironing is hard. Let me know when software can do that.
Say hello to my little sig.
If there is an algorithm for it, the Sheldon could use it to recognize (and produce?) sarcasm.
Bazynga!
She was like chocolate when she drank... semi-sweet at first and then increasingly bitter.
Wonderful, can they apply it to irony as well, because I am fucking tired of people constantly proclaiming that someone's informative statement is 'ironic'?
But sure, like that is ever going to happen.
You can't handle the truth.
The problem with that is that in American sitcoms, verbal irony is accompanied by non-verbal cues like facial expression, tone of voice, or, ugh, laugh tracks. Take away the cues, and deliver the sarcasm in a deadpan manner, and tons of people in the USA are completely unable to catch it, neurotypical or not.
Are you adequate?
That'll be good.
"The greatest lesson in life is to know that even fools are right sometimes" - Winston Churchill
No programme recognises 'sarcasm' or beauty or red (a frequency ain't 'red') All the programme does is show positive given certain sentence structures, but it don't know sarcasm from jack.
Don't we already have a tag?
There are only two types of tweets: retarded and sarcastic. If a tweet has the word "fuck" in it, it's easy to label it retarded. If a tweet has the word "retarded" in it, it's easy to label it sarcastic.
I scream. You scream. I assume that means we're both acquainted with the problem. We proceed.
"But again, we don't do it so well ourselves, do we?"
Hey, are you being sarcastic!?
Your brain is not a computer.
I have yet to meet an American who can detect sarcasm or who knows the meaning of Irony. This could seriously help you all.
Unbelievable !
Read the paper, most of what the algorithm is recognizing isn't sarcasm at all.
The examples of the classifications that are in their human curated 'Gold Standard' indicate that the curators don't know the difference between being sarcastic and simply being negative.
I suspect some SERIOUSLY fuzzy logic is involved here...