Slashdot Mirror
Toddlers May Learn Language By Data Mining
Ponca City, We Love You writes "Toddlers' brains can effortlessly do what the most powerful computers with the most sophisticated software cannot: learn language simply by hearing it used. A ground-breaking new theory postulates that young children are able to learn large groups of words rapidly by data-mining. Researchers Linda Smith and Chen Yu attempted to teach 28 children, 12 to 14 months old, six words by showing them two objects at a time on a computer monitor while two pre-recorded words were read to them. No information was given regarding which word went with which image. After viewing various combinations of words and images, however, the children were surprisingly successful at figuring out which word went with which picture. Yu and Smith say it's possible that the more words tots hear, and the more information available for any individual word, the better their brains can begin simultaneously ruling out and putting together word-object pairings, thus learning what's what. Yu says if they can identify key factors involved in this form of learning and how it can be manipulated, they might be able to make learning languages easier for children and adults. Understanding children's learning mechanisms could also further machine learning."
...I'm not quite sure it's going to change how we think about learning, as they state in TFA. I majored in linguistics, and even way back then, it was well understood by researchers in language acquisition that context played a significant role in both first and second language acquisition, but especially first. A form of data mining may well be part of the mechanics of what was happening in the experiment, but the whole way it was set up, and the way the subjects figured out what word went with what picture, had a lot to do with context. I don't mean to put down their research - this is really quite interesting - but it's also not quite the huge deal TFA seemst o suggest it is.
Data mining is just a new word for discovering statistical associations in data. Of course, children learn words by learning statistical associations between images and speech sounds; that's pretty much a tautology. I mean, what's the alternative? Divine inspiration? Toddlers running around with dictionaries?