(no title)
ryangittins | 3 years ago
I often wonder if stylometry can be used to positively identify a person based not on general word frequency, but by a single phrase or two which are rare in general but commonly used by the individual. In theory this could be relatively easy to find given a large corpus. You'd pick out the top few n-grams for short phrases by an individual and identify the ones which are most overly-represented compared to the rest of the population.
No comments yet.