Medical Linguistics

[stoutfellow]

Here's an interesting article on the use of computational linguistics to predict the likelihood of dementia - decades in advance! More study is needed, but this is definitely intriguing.

Comments

[toraks.livejournal.com]

Very interesting! Including, especially, the point that that particular measure of idea density is the one that worked the best.

I wonder how many they tried. You run into that problem with trying too many different analyses -- you increase the probability that one of them will work by chance. Or something like that.

[stoutfellow.livejournal.com]

My impression from the article is that the complexity metric was originally developed for a different purpose and then applied in this context. In any case, more studies - needed in any case - will serve to deal with the problem you raise. If the same metric (or a slight modification) continues to have good predictive value, this is big news.

Suddenly the work of Rudolf Flesch (http://en.wikipedia.org/wiki/Flesch-Kincaid_Readability_Test) leaps to mind. I'll have to think about that. (I'm not particularly fond of Mr. Flesch, but he has a lot of supporters.)

[toraks.livejournal.com]

Hmmm... after following the link, it looks like Flesch's work is what's used for saying what my reading level was through school. Funny! It always amused me to be reading many grade-levels above the grade I was in. :-)

The original article is very interesting, though. But is a bit depressing if it's true that predictors are valid that early!

[hornedhopper.livejournal.com]

This was good news to me, since I have always written in circles and at length (G). Guess I don't need to worry about *complete* dementia, then!