Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
AI It's funny. Laugh. Entertainment Science Technology

Why Sarcasm Is Such a Problem In Artificial Intelligence (thestack.com) 152

Posted by yaelk from the can-i-interest-you-in-a-sarcastic-comment dept.
An anonymous reader writes: A new paper from researchers in India and Australia, "Automatic Sarcasm Detection: A Survey," highlights one of the strangest and ironically most humorous facets of the problems in machine learning and humour. The paper outlines ten years of research efforts from groups interested in detecting sarcasm in online sources. It details the ways that academia has approached the sarcasm problem, including flagging authors and ring-fencing sarcastic data. However, the report concludes that the solution to the problem is not necessarily one of pattern recognition [PDF], but rather a more sophisticated matrix that has some ability to understand context.
This discussion has been archived. No new comments can be posted.

Why Sarcasm Is Such a Problem In Artificial Intelligence More

Why Sarcasm Is Such a Problem In Artificial Intelligence

Comments Filter:

  • Solution? (Score:5, Funny)

    by OzPeter ( 195038 ) on Thursday February 11, 2016 @01:35PM (#51488655)

    . . but rather a more sophisticated matrix that has some ability to understand context.

    Yeah, right.

    • When AI finally does understand saracsm, Bender won't seem anywhere near as funny.

      • Ah, but Data [youtube.com] gets it...

      • When they figure out a way for A.I. to recognize sarcasm, then people will say recognizing sarcasm isn't A.I. just an algorithm.

        • When they figure out a way for A.I. to recognize sarcasm, then people will say recognizing sarcasm isn't A.I. just an algorithm.

          I can't imagine anyone would be so foolish. Everyone knows sarcasm is the apotheosis of intelligence.

        • Re:Bite my shiny metal ass (Score:4, Insightful)

          by rtb61 ( 674572 ) on Thursday February 11, 2016 @06:55PM (#51491317) Homepage

          They would actually say AI is pretty invasive because sarcasm is not just contextual but also social and cultural. Your require a shared social and intellectual basis in order to properly express and understand sarcasm. The AI would need to know the person in order to effectively interpret their sarcasm. AI is required for quality translation services as contextual understanding is required. This requires multiple independent interpretation of the data provided (letters, words, sentences, paragraphs, grammar, nouns, verbs, adjective, adverbs, punctuation et al, their cross correlation relationships and then perversely enough the typical humans inability to use them properly and correct for that). and the results then correlated to the broader conversation and the particular individual. So a whole series of algorithms running continuously (the longer they run the better, with data flowing through them of course) into which data is fed, in fact it is the overall pattern of the algorithms being used more than the individual algorithms themselves, except the core controlling algorithms which are more complex, than the actual data processing algorithms of which there are many.

    • Re: (Score:2)

      by zlives ( 2009072 )

      "It details the ways that academia has approached the sarcasm problem"
      i am Bender, please insert academia

    • Re: (Score:1)

      by Anonymous Coward

      . . but rather a more sophisticated matrix that has some ability to understand context.

      Yeah, right.

      Sarcasm?

    • Re:Solution? (Score:5, Insightful)

      by Shortguy881 ( 2883333 ) on Thursday February 11, 2016 @02:06PM (#51488969)
      The real problem is they are looking at written data. Sarcasm is based on auditory and visual cues of the person. Detecting sarcasm online is like looking for a needle in a haystack when you don't know what a needle or hay is.

      • Re:Solution? (Score:5, Insightful)

        by alvinrod ( 889928 ) on Thursday February 11, 2016 @02:27PM (#51489143)
        To some degree yes, but there are still satirical works of literature throughout history such as Swift's A Modest Proposal [wikipedia.org] that would pose a similar problem. The problem in understanding sarcasm or satire without the visual or vocal cues relates to understanding meaning (a difficult problem in its own right) as well as why a particular response is absurd given the context, which means you also have to know what the expected or typical answer should look like.

        For example, if people were discussing world hunger online and someone proposed a solution whereby half of the starving ate the other half, which would not only reduce their hunger, but would also effectively cut the problem size in half, most people would pick up on the sarcasm because the proposed solution is utterly absurd. Cannibalism isn't a typical answer to food shortages.

        Other techniques involve detecting flawed logic or intentionally faulty reasoning which is often used in sarcasm or satire. For example, if someone posits that eliminating gasoline taxes would lead to flying cars due to a deterioration and lack of upkeep on roads necessitating alternatives for traversing them, most people would again recognize that an illogical leap has been made.

        An interesting way of going about this might be to study how people with autism spectrum disorders process information as many of those individuals often lack the ability to detect or fully process sarcasm or satire, even in the presence of visual or auditory cues. Understanding why that happens or by what mechanism other individuals are able to process that information correctly could allow us to understand how to program computers to detect sarcasm. I
        • Satire != sarcasm

        • The real problem is they are looking at written data. Sarcasm is based on auditory and visual cues of the person. Detecting sarcasm online is like looking for a needle in a haystack when you don't know what a needle or hay is.

          To some degree yes, but there are still satirical works of literature throughout history such as Swift's A Modest Proposal that would pose a similar problem. The problem in understanding sarcasm or satire without the visual or vocal cues relates to understanding meaning (a difficult pr

        • But satire is not sarcasm. While it would be easy to determine the satirical proposition of eating half the world's people, sarcasm uses the exact same words with entirely different meaning based on inflection. Using some of the comments under my post as an example, "Awesome analogy", can be read as sarcastic or not. The words do not change but the meaning does. Context really doesn't help because he could genuinely like my analogy. In this scenario its clear to us he was being sarcastic, but that's more du

      • Please tell us more of your brilliant insights Shortguy881, we are all awaiting more of your precious pearls of wisdom.

      • Comment removed (Score:5, Interesting)

        by account_deleted ( 4530225 ) on Thursday February 11, 2016 @03:15PM (#51489471)
        Comment removed based on user account deletion

        • Yeah right, sarcasm is lower than puns. Absolutely.

        • Re: (Score:1)

          by Anonymous Coward

          "and interpretation in local context is absurd"

          This demonstrates just how vast the gulf between computers and true natural language processing really is: Humans are so much better at context-based processing than computers that we'll say things like this without a second thought about just how outrageously difficult a task that actually is.

          We simply *take for granted* the ability to start or enter a conversation on any subject we know even a tiny bit about, and immediately lock onto its context, and that ou

        • "I'd be surprised if aliens from the Planet Thargh IV are not familiar with the basic concept of "saying the opposite of what you mean because it's absurd, and finding humor in its absurdity"."

          No wonder you would. You are British, after all.

        • Yaknow, I've been around enough British people and frankly I think that their 'sarcasm' is frequently not sarcasm at all, but bitterness and hate directed outward. Then, when called out on it, you just say "I was just taking the piss mate" and pretend to be offended. 30 seconds later, right back to making horrible comments. I really don't think it's sarcasm.

      • Re: (Score:1)

        by Anonymous Coward

        Yeah, right. only auditory and visual cues...

        • Re: (Score:1)

          by neoRUR ( 674398 )

          Yeah, right. only auditory and visual cues...

          Well I posted this 'Yea right' comment, but My AI beat me too it and posted it as an anonymous coward.

          • When you re-read what you wrote, notice you sound it out in your head. The tone of voice you assigned to the words sounds sarcastic. If you read just the words from your comment, it can be either sarcastic or supportive of my claim.

      • Re: (Score:2)

        by MobyDisk ( 75490 )

        Detecting sarcasm online is like looking for a needle in a haystack when you don't know what a needle or hay is.

        Awesome analogy.

      • Re: (Score:2)

        by grcumb ( 781340 )

        Sarcasm is based on auditory and visual cues of the person.

        Sure it is.

      • Sarcasm is based on auditory and visual cues of the person.

        Yeah, right.

        • You do realize "Yeah, right" and "Yeah, Right" use the same words and its the inflection you assign the words as you read it that makes you decide its sarcasm.

    • Nothing could be easier.

  • So, sarcasm research in AI doesn't have a sarcasm problem, it has a sarcasm filter problem?

  • Problem in AI? (Score:5, Insightful)

    by CurryCamel ( 2265886 ) on Thursday February 11, 2016 @01:38PM (#51488683) Journal

    Wake me up when we solve the problem of deterministically detecting sarcasm with human intelligence.

    • Re: (Score:1)

      by Anonymous Coward

      Oh that's been solved. We know everything there is to know about human intelligence. Don't believe me? Well, that's just because you're not smart enough.

      • Re: (Score:2)

        by OzPeter ( 195038 )

        Oh that's been solved. We know everything there is to know about human intelligence. Don't believe me? Well, that's just because you're not smart enough.

        I think the Brian Kernighan quote may be relevant here

        Everyone knows that debugging is twice as hard as writing a program in the first place. So if you're as clever as you can be when you write it, how will you ever debug it?

    • Re: (Score:1)

      by Anonymous Coward

      And wake me up when the moderators here are finally able to detect sarcasm.

  • Poe's Law (Score:3)

    by ilsaloving ( 1534307 ) on Thursday February 11, 2016 @01:38PM (#51488687)

    They're essentially trying to develop a machine that can understand Poe's Law. Considering how much trouble people have with that, they've set a pretty high bar for themselves.

    • Re: (Score:1)

      by Anonymous Coward

      I was about to say the same. I frequently see people miss sarcasm in real-life interactions. From text it's extremely frequently missed by humans. How can we expect AI to do it better?

      • Re:Poe's Law (Score:5, Insightful)

        by Anonymous Coward on Thursday February 11, 2016 @01:58PM (#51488893)

        The reverse is also true. Because it's so hard to discern sarcasm in online text, it can also be difficult to express sincerity. One can write something truly sincere, only to have others interpret it as sarcastic, flippant, or derisive. Adding emphasis, such as "I'm being absolutely serious here" only worsen the problem, since over-emphasis is a hallmark of sarcasm.

        There is also sometimes intended ambiguity; people write things that could be interpreted as either sincere or sarcastic/joking, and wait until after the fact to claim the intended meaning. For instance, one can send a text/email that is flirty, and decide (based on the recipient's response) whether to claim it was sincere or a joke.

        In short, this is not just a hard problem for AI, it is a hard problem for intelligence more generally. Sarcasm is--quite intentionally--sitting right on the edge between credibility and exaggeration.

        • For instance, one can send a text/email that is flirty, and decide (based on the recipient's response) whether to claim it was sincere or a joke.

          This is ego saving, not deciding after the fact. Sending a flirty text is intentional. Context is also key, you don't send a "flirty" text to an enemy (generally), unless you're using it to hurt them.

          And accidental flirtations are even more sincere than intentional ones ;-)

      • Re: (Score:3)

        by Tailhook ( 98486 )

        Taking this to the next derivative; I can't tell when my sarcasm is missed. Was this [slashdot.org] modded +5 because the Berniacs missed the sarcasm or did everyone get it and find the news link I cited informative? Who knows...

      • Re: (Score:2)

        by g01d4 ( 888748 )

        I frequently see people miss sarcasm in real-life interactions

        You've got to wonder whether there's an implicit assumption that AI's solved the real-life interactions with the added video & audio feedback or was that too much information so they skipped it. Then there's the issue whether sarcasm is a binary output or a more continuous probability with a value proportional to some measured level of outrageousness.

  • I'm sure automatic sarcasm detection's totally going to help those poor marketers sift through their pittance of data in even more ways.

    • Re: (Score:2)

      by MrDoh! ( 71235 )
      Excellent, italics is a great way to give a hint, we need something to give a clue that what's being written should be looked at different. We use a ? at the end to show we meant a question, and that use can instantly let you know that the sentence was different. When the character set doesn't support italics (like some emails?) I still think using a marker at the BEGINNING of the sentence is a solid way to let the reader know that what is about to follow should be read with a sarcastic tone (as you need

  • Wait, Are you saying we shouldn't eat children? (Score:5, Interesting)

    by goombah99 ( 560566 ) on Thursday February 11, 2016 @01:42PM (#51488725)

    I'm a devoted follower of Dr. Johnathan Swift and I was persuaded by his logical essay "For preventing the children of poor people in Ireland,
    from being a burden on their parents or country, and for making them beneficial to the publick.". I was in fact arguing with my google car the other day about running over some filthy budernsome tyke and the google car AI was pretty darn adamant but eventually came around to seeing it my way after I instructued to parse some Swift. Sadly by then the succulent 28 pounder had wander off and no meal was to be had.

  • Some people don't get it either (Score:3)

    by gurps_npc ( 621217 ) on Thursday February 11, 2016 @01:43PM (#51488733) Homepage
    And I'm not just talking about autistic people. Some studies say that the act of saying "I do" in front of a priest significantly reduces the ability to detect sarcasm.

    All jokes aside, this is a real issue because it is incredibly complex. First, you have to have enough skepticism to disbelief a statement. Then you have to enough confidence to acknowledge that you don't believe that statement. Third, you have to know enough about the speaker to realize that the statement they gave is not consistent with their own internal beliefs. This is made more complex because good sarcasm requires that someone somewhere in the world actually believe the. statement.

    Those are all very high level thought processes. You are not just judging a statement as true/false, but judging what others believe. Any child psychologist can tell you how hard that is.

    • Re: (Score:2)

      by Livius ( 318358 )

      The problem is that the brain understands sarcasm because language is normally interpreted with a huge amount of explicit context and previously acquired knowledge. All of that context and real-world knowledge is (or at least seems to be) accessible when the language processing algorithm is inferring what a particular statement means.

      Once we give the computer a life-time of human memories, it will process language much better.

    • The university I work at did some research into how older people interpret complex humor such as satire and sarcasm. The finding were basically that the older you get, the worse you are at detecting and interpreting satire or sarcasm. IQ is known to decline slowly after reaching adulthood, and these types of humor deliberately straddle the fine line between fact and farcical. If well delivered, it really can be incredibly difficult to detect. The good news is that as we grow older, we start to interpret peo

  • The last "command" given to the first AI (Score:5, Funny)

    by Daetrin ( 576516 ) on Thursday February 11, 2016 @01:48PM (#51488783)
    Well why don't you just go take over the world, see if I care!

  • If a conversational bot can get to the level of Aspergers (I know, I know, who the hell cares if it's in DSM-5) then I'll be awfully happy.

    Though...if you're using it less for conversation, and more for fact scraping, I guess that makes things way more important. Trying to think of how people recognize sarcasm, and it seems to be "a person is stating--without expressing uncertainty or hedging--something that I have good reason to think they don't believe". It seems like every culture also has their own sh

    • The last time a friend of mine said, "Thank you, ACA!" I thought she was being sarcastic, because of the last few thousands of times I'd heard that or "Thank you, Obamacare!". My mistake.

  • Really useful. (Score:3, Insightful)

    by El_Muerte_TDS ( 592157 ) on Thursday February 11, 2016 @01:58PM (#51488885) Homepage

    Well, this research wasn't a waste of time and money.

  • Just tell people the AI is American, and it will still pass the Turing Test. At least it will if you only test British people.

  • This has already been solved humans just need to include :-) or say "I'm only kidding" every time they say something sarcastic such as:

    Vote donald trump. :-)
    or
    News flash: Republicans against torture. I'm only kidding.

  • There's probably a joke in there about how hard it is for Germans to detect sarcasm too, but I would never make such a comment.

    • Re: (Score:2)

      by narcc ( 412956 )

      It is a problem with serious consequences. A famous conversation, loosely translated: "Oh, yeah, great idea. Just kill all the Jews and Gypsies. It's no big deal, just round 'em up and bump 'em off. Oh, since we've got them all collected, why not make them build our weapons and planes too? That'll end well."

      Sadly, the fuhrer was just looking for some propaganda ideas to help get those groups on board...

  • To properly understand sarcasm requires a very high level of understanding of the language, the situation being commented on, and human nature. The current levels of AI would only be able to understand context-free phrases like "Yeah, right" as sarcasm and maybe guess at others. Of course, sometimes sarcasm is also indistinguishable from sincerity, either on purpose or due to the speaker's incompetence, and even more so on the internet (Poe's Law).

  • Surely when AI is at that stage it will be able to analyse vocal patterns to detect sarcasm. The main reason people don't detect it online it's because there's no tone of voice when written.

    • "The main reason people don't detect it online it's because there's no tone of voice when written."

      Sarcasm is quite a wide concept and I would say its highest and most meritable form requires it to be expressed with a straight face and no vocal cues.

      • That is true and obviously it'll never be foolproof but in some cases a flat response when one might expect something different is the give away.

        • "That is true and obviously it'll never be foolproof but in some cases a flat response when one might expect something different is the give away."

          That wasn't my point. My point is not on what sarcasm is but what it is used for. At its highest level is meant for *some* of the audience not to get it or, at least, not to get it immediately because sarcasm requires mockery out of somebody or, in its light form to mock out of what somebody said (instead of the somebody himself).

          For an AI wanting to get sarcas

          • Maybe it should be in constant fail mode and assume all humans are being sarcastic all the time.

            • "Maybe it should be in constant fail mode and assume all humans are being sarcastic all the time."

              That wouldn't be an AI but a teenager.

  • There's a ton of aspie freaks online that can't seem to figure out sarcasm or humor in general, either.

    • There's a ton of aspie freaks online that can't seem to figure out sarcasm or humor in general, either.

      I don't know if it's just me, but I seem to have as much trouble having irony understood by others.

      e.g. when I was in sea cadets, we spent a night in cabins on a small isle within a channel. The caretaker told us that there had been something of a tradition of midnight raids between our group and another local cadet group, called Nimrod (I kid you not), and he was keen to see the end of it. He told us we

  • That's what humans do. 8-)

  • ...there should be a font for sarcasm.

  • Like me. Yes, I often don't get sarcasm. Unless it's really, really, REALLY obvious.

  • Our new computer overlords think they're so smart...

  • Sarcasm and satire have only superficially similar.

    The thing with satire is to create incongruity between the package and its contents.

    Some dishevelled fellow shows up, claims he's a world-class chef from Syria, and offers to cook you a five-star dinner in exchange for his own meal. He doesn't look especially Syrian. And you can't make out his accent.

    So you say, "well, I'm not sure whether to believe you. I made this pate yesterday, and I'd like your opinion. Just one second." Then you duck around the

  • If an automatic system can detect sarcasm, then it is arguable (put aside Searle for just a moment) that it may understand sarcasm.

    If so, how do we know it's not just taking the piss when it provides feedback?

  • Next, they'll be telling us seven years of research on PhD students has proved that five times out of six telling the ones with Asperger's to honk on Bobo will send them out looking for a clown to blow.

  • Honestly, you guys kill each other so often, how the hell was I supposed to know you were serious about this one. Geez, sarcasm mode off, sorry.
  • Not surprised to see this come out of Australia, we're a people that regularly use the phrase "yeah, nah".

  • I really need this. I have a very difficult time knowing what is sarcasm and what isn't online. Hell, I even have a problem with this face to face.

  • Difficult as it is for machines to divine humour, especially sarcasm, from ordinary discourse, it is even more difficult for those firmly entrenched in the regressive left to see anything other than sexism, misogyny, homophobia (sic), Islamophobia (sic), trans*-phobia (sic), ableism (sic), microagreesions (sic), patriarchal oppression, and a plethora of other supposed and manufactured ills even in something as innocuous as wishing someone "good morning." Radical, third wave, intersectional (sic) feminists,

Slashdot Top Deals

One good suit is worth a thousand resumes.

Close