Forgot your password?
typodupeerror
Image

Using AI To Identify Innuendo 86

Posted by samzenpus
from the giggity-gadgets dept.
angry tapir writes "Turning seemingly normal comments into sexual innuendo by adding the words 'That's what she said' is a cultural phenomenon. This has led some to wonder whether it is possible to determine when it is appropriate to add those magic four words to a sentence. As it turns out, identifying humor through software is hard. Two researchers at the University of Washington, however, were willing to give it their best shot. In a recently released paper entitled 'That's What She Said: Double Entendre Identification,' the researchers describe what they've found and introduce their new approach to the problem: 'Double Entendre via Noun Transfer' or DEviaNT for short." It's good to know that someone is trying to make sure the human race gets a sufficiently lewd AI one day.
This discussion has been archived. No new comments can be posted.

Using AI To Identify Innuendo

Comments Filter:
  • the BEAVIS (Binary Euphemism And Vulgar Innuendo System).
  • Innuendo... (Score:3, Funny)

    by QilessQi (2044624) on Friday April 29, 2011 @02:30PM (#35977598)

    ...isn't that a brand of suppository?

  • I would bet the computer would explode if you ran the lyrics of a Bloodhound Gang song through this software.
  • Would it label something straightforward as innuendo? For example, would the phrase "Let's have sex" be identified as having a double meaning, or would this system be able to tell that it means exactly what it says?

    • by DrEasy (559739)

      I RTFP (and not just TFA!), and they say they left that part (recognizing erotic context of the uttered sentence) for future work.

    • would this system be able to tell that it means exactly what it says?

      Well, it'll never know if it doesn't try.

      And a slap doesn't hurt a machine.

  • by Ruke (857276) on Friday April 29, 2011 @02:48PM (#35977912)
    So, do you think that these researchers are trying for an Ig Noble Award [improbable.com], or that it's pure coincidence that the primary application of their research is teaching computers to laugh at dirty jokes?
    • by godrik (1287354)

      Or you might want to add automatic "that's what she said" on the screen on tv shows. Having non-vocal "editorial" comments is very common in some program (mainly from extreme asia).

  • Great! (Score:4, Funny)

    by TempeTerra (83076) on Friday April 29, 2011 @02:56PM (#35978012)

    Next they should invent a sarcasm detector, that would be really useful.

  • I'm really frustrated with chattterbots using clever tricks to hold a conversation without understanding it. They can talk about many different things but it's clear the best of them can only respond to, at most, the previous sentence and one possible current topic. Here's an example from a chat I just had with Jabberwacky:

    Me: Would you rather be a tuna or an elephant and why?
    JW: I would much rather be a bee, and make honey for my queen.
    Me: Okay but if 'bee' weren't one of the options.
    JW: What does gender
    • by hitmark (640295)

      Sounds like the script for your average US situational comedy to me. But then i fear the AI from the article may well outperform me on the innuendo part (one aspect of being a aspie i guess).

    • by Locke2005 (849178)
      Sounds like my conversation with cleverbot. If you reference it's previous statements and point out that it contradicted itself, it just denies it said anything. It can't even make up it's mind whether it is male of female.
  • by Jane Q. Public (1010737) on Friday April 29, 2011 @03:11PM (#35978184)
    "That's What the Algorithm Thinks She Said"
  • If "that's what she said" doesn't work, "giggidy" probably does.

  • to your problem (if you know what I'm talking about)

  • As it turns out, identifying humor through software is hard.

    That's what she said.

  • Why don't the authors of this research scrape all the comments here and see if it recognizes any of the heavy breathing wordplay?

  • Am I reading the Onion?

    In a recently released paper entitled 'That's What She Said: Double Entendre Identification,' the researchers describe what they've found and introduce their new approach to the problem: 'Double Entendre via Noun Transfer' or DEviaNT for short."

    I mean ... that's just bizarre!!

    And, really, until it can master all of the places where you can use "said the actress|bishop to the bishop|actress", it will always be incomplete. =)

    Now, if you can make a computer program that can follow innue

    • I mean, the fucking use of the word fuck is in and of itself a fucking difficult thing for the fuckers to fucking figure out -- until then, they're pretty much fucked. Try hard enough, and you can make fuck into every single fucking part of speech, except maybe for those fucked up articles.

      Absofuckinglutely.
      I imagine it cannot be a simple task for any NLP to determine when the word fucking is mere fluff in a sentence, and when it is necessary to convey the meaning, and when it is necessary, WTF it stands for.

      (I know this is not very original, but I felt your post on the matter at hand was too short and could do with some enhancements. Not that the length of a post says anything about its qualities, really; short ones can be more satisfying than the longest rant if done expertly. What I mea

  • For example, the summary:

    "Turning seemingly normal comments into sexual innuendo by adding the words 'That's what she said' is a cultural phenomenon. (If you know what I mean) This has led some to wonder whether it is possible to determine when it is appropriate to add those magic four words to a sentence. As it turns out, identifying humor through software is hard. (That's what she said) Two researchers at the University of Washington, however, were willing to give it their best shot. (Yeah, I bet they are) In a recently released paper entitled 'That's What She Said: Double Entendre Identification,' the researchers describe what they've found and introduce their new approach to the problem (I'll approach your problem!): 'Double Entendre via Noun Transfer' or DEviaNT for short. (heh, short? *snicker*)"

    • by bar-agent (698856)

      "â¦and this is something that is difficult to implement in software." â" but a hardware implement will get the job done.

  • A nod's as good as a wink to a blind bat!
  • ...but let me know when it can play a game of Ar tonelico.

  • Because if you use innuendo, you must be a fucking deviant. Normal people never do it. Whoops.
  • can't wait to have my bottom patted by our naughty robot overlords.

    I hear they overclock when you play Yakkety Sax-- and you can tell them, that's what I said!

  • Maybe we'll have an app for that soon so we can finally supplement out flirting techniques.

  • They should just have asked Geoff Peterson. He's got it figured out. In your pants.

  • I think this could be really big; their task is really quite hard. First they have to suss out the meaning of the sentence, and ideally the cadence, in order to hold back until the right moment. Then they have to figure out where their addition can be legitimately inserted; not just any opening will suffice. Their biggest risk now is if they release prematurely; the timing is key and they don't want to blow it.

    I'm near Seattle in the moment, and TFA cites a presentation in Portland in June. I may just have
  • Not to get too off topic by being serious, but I'm wondering if it is even possible to detect humor just from the expression.

    Even with "that's what she said" there is an element of unpredictability that can only be tested when executed. I mean, sometimes it's not funny. In other words, the only test is if someone reads it and laughs.

    This is much like not being able to predict the outcome of code completely without executing it.

    We can always record results and rely on statistical analysis, but finding answer

  • This has led some to wonder whether it is possible to determine when it is appropriate to add those magic four words to a sentence.

    That's easy: it is. I mean, people have been determining it's (in)appropriate every time they feel the need to say it.

    Oh, you mean through software! Why didn't you say so?

You are in a maze of UUCP connections, all alike.

Working...