9
   

"Baby" Robot Learns Language Like The Real Thing

 
 
Reply Sat 23 Jun, 2012 08:39 am
"Baby" Robot Learns Language Like The Real Thing
June 23, 2012
by Dana Farrington - NPR

A baby robot has been born. Now, with little DeeChee's help, researchers are studying how babies transition from babbling to forming words.

Dr. Caroline Lyons of the University of Hertfordshire is one of the computer scientists who helped design DeeChee the robot. She tells Weekend Edition host Scott Simon that humans are also critical to their experiments.

"One of our ideas is that you only learn to speak by interacting with another human. So we brought in a lot of ordinary participants and just asked them to teach DeeChee the names of shapes and colors."

As one might expect, not everyone is a great teacher. The study, published by the Public Library of Science One, puts it diplomatically:

"Some participants were better teachers than others: some of the less good produced very sparse utterances, while other talkative participants praised DeeChee, whatever it did, which skewed the learning process towards non-words."

Along with the study, researchers posted a video of a participant teaching DeeChee.

Lyons says the experiments, which last just eight minutes, are intended to simulate a focused, therapeutic learning session, rather than the gradual learning babies do in their everyday environments.

DeeChee started out knowing most of the syllables in the English language, as Slate reports.

"As humans talk, DeeChee tracks the number of times different syllables are used. It then uses the more common sounds to recognize words, which it can then speak."

Human baby Charlotte, the 13-month-old daughter of NPR producer Tom Bullock, tried the same tests that DeeChee, the robot, does for language-learning experiments. Dr. Caroline Lyons says human babies have an advantage: They spend every waking hour of the day in a speaking world.

The robot keeps a record of the syllables it hears and speaks, producing data for the researchers.

DeeChee has white plastic skin and a smile of red lights. Its hands can grab and gesture — and the participants respond to its human-like features.

"You can see how people do enter into the spirit of the thing," Lyons says, "and they do tend to talk to DeeChee as if it was a small child."

Wired Magazine explains that the life-like design of humanoid robots like DeeChee is not just for the participants' sake.

"Many researchers think certain cognitive processes are shaped by the bodies in which they occur. A brain in a vat would think and learn very differently than a brain in a body."

Photos and voices

http://www.npr.org/blogs/alltechconsidered/
 
fresco
 
  1  
Reply Sat 23 Jun, 2012 11:29 am
@BumbleBeeBoogie,
University of Hertfordshire !.... Laughing
With respect (to the good Doctor), no wonder there is "criticism".
0 Replies
 
Setanta
 
  2  
Reply Sat 23 Jun, 2012 11:58 am
@BumbleBeeBoogie,
Aunt Bee's source wrote:
"You can see how people do enter into the spirit of the thing," Lyons says, "and they do tend to talk to DeeChee as if it was a small child."


Why would one speak differently to a small child than a adult? I've kept infants quiet and completely enthralled for long periods of time by speaking to them in the same way i would to an adult. They know they're being spoken to seriously, they can tell the difference between that and the goo-goo ga-ga baby talk ****. If one objects that a small child hasn't the vocabulary to understand, i ask how they are supposed to acquire it.
fresco
 
  1  
Reply Sat 23 Jun, 2012 12:06 pm
@Setanta,
Convergence to "child language" by changing intonation patterns and syntax is a well documented "instinctive behaviour" demonstrated by carers of infants. The significance of this behaviour is not fully understood, but it may have something to do with the presence of "mirror motor neurones" found in many species, whose staged operation may be part of the maturation process.
Setanta
 
  1  
Reply Sat 23 Jun, 2012 01:58 pm
@fresco,
Is "carers" supposed to mean those who care for infants? The groves of Academe would wither away without jargon.
sozobe
 
  1  
Reply Sat 23 Jun, 2012 04:11 pm
@Setanta,
"Motherese" has been proven to be very helpful for picking up a language, actually.

From "The Scientist in the Crib" by Alison Gopnik (quoting a swath but the last bit is most important):

Quote:
The tests show that babies' preferences have nothing to do with the actual words mothers use. Babies choose motherese (or "parentese" or "caretakerese") even when the speaker is talking in a foreign language so infants can't understand the words, or when the words have been filtered out using computer techniques and only the pitch of the voice remains. Apparently they choose motherese not just because it's how their mother talks but because they like the way it sounds. Motherese is a sort of comfort language; it's like aural macaroni and cheese. Even grown-ups like it. Pat's graduate students discovered that listening to the lab tapes of motherese in a foreign language was a wonderful therapy for end-of-term stress. The mother's voice is an acoustic hook for the babies. It captures babies' attention and focuses it on the person who is talking to them.

The elaborate techniques of computer voice analysis reveal exactly what it is we do when we talk to an infant. The pitch of our voice rises dramatically, sometimes by more than an octave; our intonation becomes very melodic and singsongy; and our speech slows down and has exaggerated, lengthened vowels.

Motherese is a universal language. People across all cultures do it when they talk to their infants, even though they usually aren't aware of doing it at all. When mothers listen to recordings of themselves producing motherese, the reaction is: That can't be me. I sound really stupid. Should I be doing that? But they do it intuitively, without conscious awareness.

Why do we do it? Do we produce motherese simply to get the babies' attention? (It certainly does that.) Do we do it just to convey affection and comfort? Or does motherese have a more focused purpose? It turns out that motherese is more than just a sweet siren song we use to draw our babies to us. Motherese seems to actually help babies solve the Language problem.

Motherese sentences are shorter and simpler than sentences directed at adults. Moreover, grown-ups speaking to babies often repeat the same thing over and over with slight variations. ("You are a pretty girl, aren't you? Aren't you a pretty girl? Pretty, pretty girl.") These characteristics of motherese may help children to figure out the words and grammar of their language.

But the clearest evidence that motherese helps babies learn comes from studies of the sounds of motherese. Recent studies show that the well-formed, elongated consonants and vowels of motherese are particularly clear examples of speech sounds. Mothers and other caregivers are teachers as well as lovers. Completely unconsciously they produce sounds more clearly and pronounce them more accurately when they talk to babies than when they talk to other adults. When mothers say the word bead to an adult, it's produced in a fraction of a second and it's a bit sloppy. But when mothers say that same word to their infants, it becomes beeeeeed, a well-produced, clearly articulated word. This makes it easier for infants to map the sounds we use in language.


Quote found here:

http://memexplex.com/Meme/906/

In context (but not copy-and-paste-able) here:

http://books.google.com/books?id=ui6KAniUJfsC&pg=PA130&lpg=PA130&dq=alison+gopnik+motherese&source=bl&ots=xioXV6UBFA&sig=Cs4tR4VFroymBmi-NB2SC9kiQaE&hl=en&sa=X&ei=ST7mT8CZC4u66AH38NDgDg&ved=0CFkQ6AEwBQ#v=onepage&q&f=false
Setanta
 
  1  
Reply Sat 23 Jun, 2012 04:12 pm
@sozobe,
I don't know why you're addressing this to me.
sozobe
 
  1  
Reply Sat 23 Jun, 2012 04:14 pm
@Setanta,
Setanta wrote:
Why would one speak differently to a small child than a adult? I've kept infants quiet and completely enthralled for long periods of time by speaking to them in the same way i would to an adult. They know they're being spoken to seriously, they can tell the difference between that and the goo-goo ga-ga baby talk ****. If one objects that a small child hasn't the vocabulary to understand, i ask how they are supposed to acquire it.


(Emphasis mine.)

What I quoted above from Alison Gopnik is why one would speak differently to a small child than an adult. Especially if the purpose is expressly for the robot to pick up language in a way parallel to how human children pick up language.
Setanta
 
  1  
Reply Sat 23 Jun, 2012 04:25 pm
@sozobe,
What you have posted with it's jargon (motherese) provides no evidence that the child prefers to be spoken to in that manner. I can see that they might be comforted to hear reassuring noises from a mother. That is not, however, evidence that that is the only way, the best way, or even the preferable way for children to learn language.

Of course, i was also recounting my anecdotal experience, and didn't claim it was anything else.

I don't know why you're addressing your remarks to me.
sozobe
 
  1  
Reply Sat 23 Jun, 2012 04:28 pm
@Setanta,
Because I see the "goo-goo ga-ga baby talk ****" sort of comments a lot, and one thing I found surprising as a mom was how I went to that fairly automatically, and one thing I found interesting in that book was the explanation of why we do that and why it's a good thing.

People often assume it's unnecessarily dumb/ cutesy, but it's actually a vital part of learning a language.

You can read more of the book in the Google link if you'd like, it explains more about why it is a very effective way for children to learn language.
Setanta
 
  1  
Reply Sat 23 Jun, 2012 04:31 pm
@sozobe,
How unfortunate that you were offended, I will emphasize one again that i was recounting my anecdotal experience, and made no claims for that. I don't know why you're addressing your remarks to me.
snood
 
  5  
Reply Sat 23 Jun, 2012 04:43 pm
Do you ever take a day off from being a nasty pain in the ass, Setanta?
0 Replies
 
Setanta
 
  1  
Reply Sat 23 Jun, 2012 04:44 pm
From the wikipedia article on "motherese."

Quote:
Shore and other researchers believe that baby talk contributes to mental development, as it helps teach the child the basic function and structure of language. Studies have found that responding to an infant's babble with meaningless babble aids the infant's development; while the babble has no logical meaning, the verbal interaction demonstrates to the child the bidirectional nature of speech, and the importance of verbal feedback. Some experts advise that parents should not talk to infants and young children solely in baby talk, but should integrate some normal adult speech as well. The high-pitched sound of motherese gives it special acoustic qualities which may appeal to the infant. Motherese may aid a child in the acquisition and/or comprehension of language-particular rules which are otherwise unpredictable; an example is the reduction or avoidance of pronoun reversal errors. It has been also suggested that motherese is crucial for children to acquire the ability to ask questions. Some[who?] feel that parents should refer to the child and others by their names only (no pronouns, e.g., he, I, or you), to avoid confusing infants who have yet to form an identity independent from their parents.


Note the conditional nature of these claims. Note that there are no references to any control. Is there any meaningful data on language learning in the absence of "baby talk?' Note that nothing precludes the child from learning the bidirectional nature of speech from being addressed in a normal, conversational manner. Note that some alleged experts say that infants should not be addressed in baby talk only.

I find the claims for motherese to be suspect at best. Without a control, what quality of information can be derived from the putative studies? All of which, combined with the anecdotal nature o my comments leads me to wonder why you were addressing me.
sozobe
 
  1  
Reply Sat 23 Jun, 2012 04:44 pm
@Setanta,
Where do you see offense?

Yours is a common misconception, which I'm commenting on and providing some evidence as to why it's a misconception.

Which happens to be quite pertinent to the thread. (Speaking to a "baby" robot as one would to a human baby is more likely to result in the "baby" robot successfully learning language.)
Setanta
 
  1  
Reply Sat 23 Jun, 2012 04:49 pm
@sozobe,
You haven't demonstrated any misconception. Read the Wikipedia article, the section on universality, which contains counter claims that baby talk is not universal, and refers to Samoan culture, wherein adults don't talk to their infants at all.

That i disagree with you is not evidence that i am laboring under any misconception. You have offered absolutely no evidence that this would help "robot babies" learn language. You haven't established the truth of your claims.
ossobuco
 
  1  
Reply Sat 23 Jun, 2012 04:49 pm
Interesting subject. I'll buy (for now) the data on motherese and Fresco's comment, but I also get Set in that I never have used motherese, my not being in that situation. I met my niece early, but wasn't around her much until she was two. I always talked Osso to her, for better or worse. We communicated seriously virtually immediately and that goes on and on.

Meantime, I may sometimes speak a version of motherese to my dog. Probably not, but it veers close.
0 Replies
 
sozobe
 
  1  
Reply Sat 23 Jun, 2012 05:17 pm
@Setanta,
Setanta wrote:
You have offered absolutely no evidence that this would help "robot babies" learn language.


I have, actually. A link to a book that goes into this in great detail. I found it convincing.

This thread is about teaching a robot to learn language in the same way a human learns language. If a human's language learning is facilitated by motherese, it makes sense that this robot's language learning would be similarly facilitated.

At any rate, you now seem to get why I would address my comments to you. If you disagree about the utility of motherse, that's fine, and a debate about it might be interesting -- I have limited time for that right now. Most of what I would say can be found in "The Scientist in the Crib" anyway.
sozobe
 
  3  
Reply Sat 23 Jun, 2012 05:22 pm
@sozobe,
I did just quickly check out your link, though, and right at the beginning there is this:

Quote:
Baby talk is more effective than regular speech in getting an infant's attention. Studies have shown that infants actually prefer to listen to this type of speech.[8] Some researchers, including Rima Shore,[9] believe that baby talk is an important part of the emotional bonding process between the parents and their child that help the infants learn the language. Other researchers from Carnegie Mellon University and the University of Wisconsin confirm that using basic “baby talk” helps babies pick up words faster than usual.[10] Infants actually pay more attention when parents use infant-directed language, which has a slower and more repetitive tone than used in regular conversation. This child-directed speech has also been shown in languages other than English.[11]


(Emphases mine, all numbers link to studies in the article.)
Thomas
 
  3  
Reply Sat 23 Jun, 2012 08:19 pm
@sozobe,
sozobe wrote:
Because I see the "goo-goo ga-ga baby talk ****" sort of comments a lot, and one thing I found surprising as a mom was how I went to that fairly automatically, and one thing I found interesting in that book was the explanation of why we do that and why it's a good thing.

Interesting. Now you've got me wondering how motherese looks like in ASL, and if that comes naturally to you, too.
dlowan
 
  1  
Reply Sat 23 Jun, 2012 08:44 pm
@Thomas,
I suspect she speaks it like a native.
 

Related Topics

New Propulsion, the "EM Drive" - Question by TomTomBinks
The Science Thread - Discussion by Wilso
Why do people deny evolution? - Question by JimmyJ
Are we alone in the universe? - Discussion by Jpsy
Fake Science Journals - Discussion by rosborne979
Controvertial "Proof" of Multiverse! - Discussion by littlek
 
  1. Forums
  2. » "Baby" Robot Learns Language Like The Real Thing
Copyright © 2024 MadLab, LLC :: Terms of Service :: Privacy Policy :: Page generated in 0.03 seconds on 04/26/2024 at 10:30:44