Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Government Programming Software The Internet United States Idle Technology

US Metaphor-Recognizing Software System Starts Humming 105

coondoggie writes "An innovative project, called Autonomous Dynamic Analysis of Metaphor and Analogy, or ADAMA, aims to build a software system that can automatically analyze metaphorical speech in five different languages by analyzing huge quantities of online data got off the ground this week when the U.S. Army Research Laboratory awarded a $1.4 million contract to the team conducting the research. The research is backed by the US Intelligence Advanced Research Projects Activity (IARPA), which develops high-risk, reward research projects for the government, and is intended to build a repository of speech metaphors from American/English Iranian Farsi, Mexican Spanish and Russian speakers. ADAMA could have immediate applications in forensics, intelligence analysis, business intelligence, sociological research and communication studies, researchers stated."
This discussion has been archived. No new comments can be posted.

US Metaphor-Recognizing Software System Starts Humming

Comments Filter:
  • The First Post is a metaphor for testing the beginnings of censorship.

    • by swx2 ( 2632091 )
      so is that meta-first?
      • by s.petry ( 762400 )

        The problem as someone points out below is that metaphor is brute force, not natural language. It can't be done naturally. Here is an example: If I said "Installing your code is like gently pulling teeth with a screw driver" most people would get the metaphor. It's not common, so has to be brute forced into a search. Now if I change "screw driver" to "gentle tug" there is a whole new meaning.. and the Metaphor dictionary would have to be updated.

        Metaphor is constantly changing, and there is no way to i

        • I thought a "Lewinsky" involved a cigar, which most definitions of "BJ" do not.
          • Sometimes a cigar is not a cigar.

          • by s.petry ( 762400 )

            There was a different metaphor for the cigar being a Lewinski. Again, one of the many problems trying to build AI to handle metaphor. Language is very fluid, context sensitive, and tone sensitive. You can not build AI to do this, given what we have now at least.

            Next, I guess we will have to pass laws that restrict all new metaphors and variations from "approved" language.

            One thing I did not point out is that to do this as an anti-terrorist measure as it's being masked, you need to have the same massive d

            • Well I think "can't build" is too strong a phrase. You sure may be able to produce some very good aproximation of an AI that can handle metaphor with a huge database.

        • So this is why they allowed warrantless wiretapping. So they could build a huge database of voice data.

          At least it makes sense now.

        • by swx2 ( 2632091 )
          Well, i don't think just building a repository will be too hard. The AI doesn't have to understand it, it just has to be there stored somewhere. Maybe you can get something like reCaptcha for the analysis part?
        • Are you maybe confusing metaphor with idiom? Idiomatic speech could truthfully be impossible to interpret without knowledge. A metaphor OTOH is a type of analogy. It operates on the basis of similarity or likeness with something else. It is thus amenable to logical analysis. Take your example, would someone who has not heard this example before understand it? I think it is self evident the answer is yes. It is therefor proven that this metaphor is open to computational analysis.

          As language changes some meta

          • by s.petry ( 762400 )

            Idiom would have to be included, but that is not my main point of reference.

            Let me simplify the thought a bit, and take out all of the computer jargon for a minute.

            Humans that natively speak language can not properly map the language. There is no circuit that can be built to do so, hence why we have so many Linguists and Language professors and students. Look at the complexities in language rules (let alone metaphor) between say Texas and Virginia. We can not properly map those out, and without the abili

            • I think the assumption that there's anything impossible about doing it is what is wrong. I also think we are definitely reaching the levels of processing power necessary to meaningfully attack these sorts of problems. Nothing will every do it exactly like a human, but the path to doing it better than the best human is clearly open ahead of us. I can see a world where you can have EVERYTHING parsed for you into an ideal form for you particularly to understand it best. Is the govt going to get a super-human s

        • : If I said "Installing your code is like gently pulling teeth with a screw driver" most people would get the metaphor. >

          Simile not a metaphor. Using 'like' is a simile. You mean 'i enjoy the sensation of pulled teeth so much i will happily install your code!'

        • If I said "Installing your code is like gently pulling teeth with a screw driver" most people would get the metaphor

          Well, except that's simile [wikipedia.org] not a metaphor.

          How fast for example did the metaphor for a BJ change to and from "Lewinski"?

          That's a euphemism [wikipedia.org]. Again, not a metaphor.

          So hopefully the people writing this remember a little more of high school literature. ;-)

        • No, sorry, it is really much easier than that.

          It can be done by "emotional conversions".

          "Installing your code is (Something from the list of things that are hard)."

          So in the list of things that are hard are spending a day with your mother in law, changing a tire in the rain at night, driving with one of your gears missing. Anything from that category works. It's like Mad libs.

          Give 12 college kids 5 pizzas and 8 liters of cola for three days and you could have a list of 100,000 of them.

  • Every time a task like this is mastered it's suddenly not considered human level intelligence anymore. I can't believe it's long now...
    • Indeed.

      The only thing slowing this down is a thundering existential issue unlike anything we've ever seen. Suddenly we're gonna need to be reviewing the "Science Fiction" section of the lit world for some advice how to handle the rise of sentient AI.

      And notice it's *only* 1.4 million. That's all? That's like a 4 person team plus a building plus equipment. It's the "Damn with faint praise" department - it's so it can make news but purposely underfund it for external meta reasons.

      We would have had hard AI alr

      • by lennier ( 44736 )

        Suddenly we're gonna need to be reviewing the "Science Fiction" section of the lit world for some advice how to handle the rise of sentient AI.

        Kubrick 1968 suggested "screwdriver" while Cameron, 1984 gave a persuasive argument in favour of "shotguns and homemade pipe bombs".

        We would have had hard AI already if we had spent the last 10 years and the trillion dollars on it like we pushed for the moon. But no, Beating People Up is more fun.

        Who's to say you can't have both? That's why it's called the Defense Advanced Projects Agency!

        But seriously, DARPA did in fact sponsor pretty much exactly the kind of "moon shot" program you are arguing for, throughout the 1980s. They spent a billion dollars and it was called the Strategic Computing Initiative [wikipedia.org] (as a marketing phrase to compete with Strategic Defense Initiative

        • Google, Cyc, Wikipedia, and Watson are damned impressive. I wouldn't mind the government spending a billion dollars making another set possible
        • In fact, those tools are perhaps part of the way a next gen AI would emerge. I'd heard briefly of the AI winter, but the hardware is different now and those other low-level tools are part of the key.

          Of course it's currently cheaper to get people from emerging countries to do certain things - however I forsee the next AI as chained Expert modules plus "middleware". Remember the Loebner challenges? Until a few years ago (if at all) the contestants went in looking to sniff out the computer programs with a deli

      • we did not spend trillions pushing for the moon, mere hundreds of billions scaled to today's dollar. Over ten trillion on our nuclear arsenal, yes.
    • by Frogg ( 27033 )
      agreed: slowly but surely a.i. is getting there :)
    • by nbauman ( 624611 )

      Actually, it's not artificial intelligence. It's cheap computers. Peter Norvig said that people were trying to write programs that would understand language, so they could translate it. It didn't work. At Google, they gave up on trying to understand the language, and figured out how to do it with brute-force algorithms. No need to understand what you're translating. http://www.stonetemple.com/search-algorithms-with-google-director-of-research-peter-norvig/ [stonetemple.com]

      Most of what you read is nonsense anyway. http://nor [norvig.com]

    • Every time a task like this is mastered it's suddenly not considered human level intelligence anymore. I can't believe it's long now...

      Oh please. It's better than human intelligence. A computer will recognize "Man, that's the bomb!" in the context of a conversation as a statement of approval or respect. A person would dial 911, then shoot the guy 16 times for waving a suspicious looking bag of candy. -_-

      • "A TSA agent would dial 911, then shoot the guy 16 times for waving a suspicious looking bag of candy."

        There, fixed that for you.

  • ...a kick in the pants!

  • by flibbidyfloo ( 451053 ) on Wednesday May 09, 2012 @02:19PM (#39945553)

    I bet that computer is going to be busier than a one-legged humanoid robot in an ass-kicking contest.

  • I imagine this would be extremely useful to recognize and block new ways of referencing forbidden topics in countries that censor the internet and text messaging.
    • by Strawser ( 22927 )

      Or for scanning social media, as our (US) government does ( http://yro.slashdot.org/story/12/02/29/144257/what-the-dhs-is-looking-for-in-your-posts [slashdot.org] ). I guess scanning for specific keywords isn't particularly useful since no one actually uses those keywords. "Stick it to the man" (or whatever) is more likely than "use high explosives to target key government infrastructure, such as bridges and airports" (or whatever).

      Seriously, who's going to tweet something like that?

      • by JLDohm ( 741501 )

        Seriously, who's going to tweet something like that?

        or post it on the internet...

        I suppose it would be useful for scanning social media, but I still have a hard time believing that scanning social media will ever be useful. Too many false positives.

        • by Strawser ( 22927 )

          I suppose it would be useful for scanning social media, but I still have a hard time believing that scanning social media will ever be useful. Too many false positives.

          I think they would ultimately be looking less for specific threats than trying to profile individuals for follow-up snooping. Ie.: Someone who's politically extreme, associated with radical groups, uses aggressive language, obsesses over politics, FB "Likes" survivalist sites, extremist groups, etc. -- just to collect individuals names' to later use more focused forms of snooping. This kind of DB would be useful in spotting "problematic" demographic groups.

          • by JLDohm ( 741501 )
            I suppose the sum of many individual communications could make such a system work. A pattern of communications could provide enough specificity to target a small enough number of individuals that would make more targeted snooping feasible.
            • by Strawser ( 22927 )

              Yeah, that's what I was thinking. Take the existing social media scanning code, mix it with this type of language processing code, and let a group of psychiatric professionals define "problematic" personality profiles to build some kind of scoring system for. Like an MBTI for terrorists (or whoever the latest boogie man is, once they have it in place).

      • by c0lo ( 1497653 )

        Seriously, who's going to tweet something like that?

        In the context of metaphors, some clueless UK [slashdot.org] tourists?

  • by DanTheStone ( 1212500 ) on Wednesday May 09, 2012 @02:23PM (#39945609)
    Sometimes punctuation is important. "American/English Iranian Farsi, Mexican Spanish and Russian speakers" doesn't make any sense.
    • It makes perfect sense: This program is written for speakers of American Iranian Farsi, English Iranian Farsi, Mexican Spanish, and Mexican Russian.

      At least, that's what most NLP would probably make of it.

  • I assume it's used in conjunction with the

    Generic Automatic LAnguage Connector and Translator Improving Context Analysis (otherwise known as GALACTICA)?

    • FUND IT!

      Sadly, it'll face constant harassment by the Counter-Yelling Language Obscuring Nodes. Chat bots that yell obnoxiously at people using Frankenstein mash-ups of language colloquialisms. Of course this is slashdot, so it'll be hard to tell them apart from humans.
  • why is this story categorised as 'idle'?? it's a perfectly valid news story about artificial-intelligence/machine-learning & big-data, the kind of story that is becoming more and more common as time progresses -- this is the future!

    i've added the tag 'bigdata' to the story more than once, but it keeps getting removed -- why are people so clueless?

    i even logged for the first time in years, just so i could set the correct tags. meh.

  • by Translation Error ( 1176675 ) on Wednesday May 09, 2012 @02:29PM (#39945691)
    And what software wrote that summary? Ouch, that thing was painful to read. And I have no idea why the title says the software has 'started humming'. Doesn't anyone go over these things to make them a bit more readable before putting them on Slashdot's front page?
    • Doesn't anyone go over these things to make them a bit more readable before putting them on Slashdot's front page?

      Of course not. Are you new here?

    • And I have no idea why the title says the software has 'started humming'.

      Daiiiisy.
      Daaaiiiiiisy.

  • I once knew a guy who never meta4 he didn't like.

  • by Anonymous Coward

    Farsi is what they call their language. We call it Persian. We don't refer to German as Deutsch do we? Also, there is little use in specifying that it is Iranian Farsi, as the dialects within the country vary as widely as between the different countries that use the language.

  • by Jeremi ( 14640 ) on Wednesday May 09, 2012 @02:37PM (#39945797) Homepage

    ... and as is so often the case, Douglas Adams predicted it decades in advance:

    ------

    "Well," [Richard] said, "it's to do with the project which first made the software incarnation of the company profitable. It was called Reason, and in its own way it was sensational."

    "What was it?"

    "Well, it was a kind of back-to-front program. It's funny how many of the best ideas are just an old idea back-to-front. You see there have already been several programs written that help you to arrive at decisions by properly ordering and analysing all the relevant facts so that they then point naturally towards the right decision. The drawback with these is that the decision which all the properly ordered and analysed facts point to is not necessarily the one you want."

    "Yeeess ..." said Reg's voice from the kitchen.

    "Well, Gordon's great insight was to design a program which allowed you to specify in advance what decision you wished it to reach, and only then to give it all the facts. The program's task, which it was able to accomplish with consumate ease, was simply to construct a plausible series of logical-sounding steps to connect the premises with the conclusion.

    "And I have to say that it worked brilliantly. Gordon was able to buy himself a Porsche almost immediately despite being completely broke and a hopeless driver. Even his bank manager was unable to find fault with his reasoning. Even when Gordon wrote it off three weeks later."

    "Heavens. And did the program sell very well?"

    "No. We never sold a single copy."

    "You astonish me. It sounds like a real winner to me."

    "It was," said Richard hesitantly. "The entire project was bought up, lock, stock, and barrel, by the Pentagon. The deal put WayForward on a very sound financial foundation. Its moral foundation, on the other hand, is not something I would want to trust my weight to. I've recently been analysing a lot of the arguments put forward in favour of the Star Wars project, and if you know what you're looking for, the pattern of the algorithms is very clear.

    "So much so, in fact, that looking at Pentagon policies over the last couple of years, I think I can be fairly sure that the US Navy is using version 2.00 of the program, while the Air Force for some reason only has the beta-test version of 1.5. Odd that."

    "Do you have a copy?"

    "Certainly not," said Richard. "I wouldn't have anything to do with it."

  • You have to roll the hard six.
  • Pointing down... (Score:4, Interesting)

    by bennomatic ( 691188 ) on Wednesday May 09, 2012 @02:49PM (#39945961) Homepage
    I've got your metaphor analysis system *right here*!

    Personally, I think such a system should be called the "Innuendo Engine" as sexual references would end up being the underlying context for the vast majority of decyphered metaphors.
    • Say no more... say no more! wink wink nudge nudge

    • by Anonymous Coward

      I've got your metaphor analysis system *right here*!

      Personally, I think such a system should be called the "Innuendo Engine" as sexual references would end up being the underlying context for the vast majority of decyphered metaphors.

      Oh goodness, all anyone here seems to think about is innuendo and worrying about the size of their metaphor. We're obsessed. We see a girl come on here, and maybe she's got a nice simile or she's really witty, but all everyone seem to care about is the size of her devices, whether they could get her to give them a euphemism, and imagining if she'd be up for a double entendre.

  • When you write a story about an AI system and say it is "humming" then by God, the story had better contain a YouTube video of the AI humming a Beatles tune or something through a tinny PC speaker.

    Merely being turned on does not mean it's singing.

  • Holy crap, we are well on our way to Skynet and Cylons. Too bad we all can't escape to the 13th colony because we're already on it.

    • no worries, it's all cyclical. this 13th colony also will someday be the 26th, which is to say the future 13th
  • I'd love to see how this handles Shakespeare. Or Joyce, for that matter. Also, it would be really cool if you could translate to idiomatic metaphors that make sense in the target language.
  • Well that's as slick as a baby sliding across a greasy, frozen over pond in January. Chew on that metaphor Mr Algorith!
  • A computer with metaphor-recognizing software is like a... wait, sorry, I was thinking of simile-recognizing software.

  • Google Translate: English Cockney

  • ...the Contemporary Youths' Lexicon Of Neologisms!
  • hey duds wanna analyze abstractions for meaning, sure dude, as soon as we get mind reading technology to determine "What were they thinking?" and "what meaning did they attach?" I.e. Political double speak.... the bottomless pit of trolling abstraction usage.

Truly simple systems... require infinite testing. -- Norman Augustine

Working...