Based on my experiments (not as extensive as yours), I think the gizmo is -- and this isn't actually surprising -- focusing much more on vocabulary than it is on "style" as such. Running word-usage comparisons against a database of textual samples strikes me as not too complex from a codewrangling perspective; OTOH, running the same word-usage comparisons plus routines that would check for things like verb tense, sentence structure, narrative point of view, etc. strikes me as way complicated.
no subject
Date: 2010-07-13 07:44 am (UTC)