Sunday, November 13, 2016

Developing language – should tech learn as kids learn?

In the readings this week we learned about how children comprehend phrases differently. Studies the acquisition of scalar implicature occurs later in life, leaving children with these different understandings of phrases. Scalar implicature occurs in four phases. The first phase usually works the same for both children and adults in that we often come up with the same literal meaning. However, children differ in the final three stages by generating different or fewer alternate sentences, restricting different or fewer alternatives that are not as informative, and not strengthening their interpretations. Some examples of how children interpret things differently include how they interpret sentences using ‘and’ and ‘or,’ ‘must’ and ‘might,’ and ‘all’ and ‘some.’ The lines between these blur when children respond to questions using these words. They will treat ‘or’ as ‘and’ and will accept ‘some’ as a descriptor when ‘all’ is more accurate. However, adults use counterfactual theories for pragmatic inference when they do not have enough information. This (lack of information) often balances out the differences between children and adult understanding. Possessing a feature is determined to be less common than not possessing a feature so adults will counter choosing some photos when asked which a phrase describes.


I find it super interesting that children develop from not distinguishing these words and patterns to accepting them as having entirely different meanings. I would like to look into more research on how this development takes place. This information could theoretically help develop artificially intelligent beings. We could start simple and create a program that does not distinguish these terms. Then, we could try to implement learning into the program and make the machine smarter. Learning would model a child’s brain and hopefully develop its own learning skills to understand language as, and no more than, humans do. Even if not for machine learning, a better understanding of development would help developers create a better functioning device that communicates well with humans.

No comments:

Post a Comment