I saw this mentioned in a science magazine some time ago, and recall that the algorithm looks for words and phrases coded in English for interpersonal warmth, coziness, and casualness, and weights these as feminine. The male weighting is for less intimate locutions and 'drier' phraseology. Since reading that I have myself been guessing at the authorship of New York Times articles and other material. I've been right about 75% of the time I think, maybe better.
Allen
|