A recent study notes with alarm that young people are spending less time reading and more time on the internet -- where the primary form of interaction is: Reading
Hmmm. Reading counts for less because it's on a screen than on a printed page?
Let's face it. If you can't read fairly well you're going to have trouble using the internet, even to play games. And as we all know reading is a skill that improves with practice.
So logically. . .
But this is the evil old internet. Good things _can't_ be coming out of it.
Subscribe to:
Post Comments (Atom)
6 comments:
Generally, I agree, but I do note that in certain Internet communities, the *quality* of writing is atrocious, because so many of the would-be writers have not been exposed to nearly as much literate writing, and so they don't recognize their lack of coherence.
Or maybe I'm just too much of an old fart, and I can't handle change...
Yes, writing is a different story. Part of that is that our language is changing at an unusually rapid rate just now because American English has become a world language, spoken and written by millions of people for whom it is, at best, a second language.
Part of it is the truly abysmal state of our school systems.
And part of it is the lack of a semantic spell checker for the masses.
I'm way behind on language processing research, but I wasn't aware of anything even close to viable in the area of semantic validation.
I do remember reading about research on Augmented Transition Networks, which added annotations to the parse trees that allowed for things like subject/verb matching as far as plurals, but nothing that would cover the kind of language mangling I see on the on-line fiction sites.
Is Cyc trying to do anything like this, or are they still in database-construction mode?
Combining your points about second languages and inadequate native educational systems, I do note that European posters seem to have much better English-language skills, even though they're more likely to apologize for not being a native speaker.
AFIK we're a long way from having a true semantic web, especially if we approach it from a purely theoretical standpoint.
However I think we are going to get something useful much more quickly. It won't be perfect semantic understanding, but it will be good enough to get 80-90 percent of it right.
Of course the problem with that is that spell checkers are in general 80-90 percent accurate. And you can see what a mess that can produce.
Some of the failures I see at on-line fiction sites are easy to imagine being fixed automatically. For example, their/there/they're homophones and pseudo-homophones caused by spelling so bad the 'valid word' checker guesses wrong when suggesting a correction 'Capitan' for "Captain" for example.
Other errors are beyond anything short of Citizen-level AI, I think. Run-on sentences you can trap since they're rarely grammatically correct, but half-page paragraphs would be trickier.
Almost any improvement would be nice, though.
Part of the problem with online fiction sites is that fiction writers of all levels of ability are, in general, lousy spellers.
Trust me on this. :-)
That said, homophone problems are rampant on the internet. A simple fix would be to have spell checkers display short definitions with the words when they offer alternatives.
What I think is likely to bring us functionally close to the semantic web is a wikified neural net or something similar -- a technique that combines human input from a lot of readers with a some method of "learning" what the writer most probably means.
Note I said "functionally close". I am not at all sure such a technique could be expanded to produce a theoretically correct semantic web. But frankly I don't care, if it will help the level of writing.
Post a Comment