On 12/15/05, SJ 2.718281828@gmail.com wrote:
Yes, this makes it rather hard to design a good blinded study. Feeding all text through a dialect-simulator, or better yet an editor that unifies paragraph-length and word choice without altering any facts or much clarity, might help. The design of a decent reference-work comparison study might merit a paper in itself.
SJ
A program that can edit an article to remove all the verbage and keep the facts is pretty far away, from an AI stand-point... If they had had the time and money, though, I guess they could have had their students read through the articles and simply list each "fact". Then you could get the number of errors per 100 facts. Not quite as easy, though.
Sam
-- Asbestos http://en.wikipedia.org/wiki/User:Asbestos