Kevin Drum reports that “the internet toy du jour is a site that supposedly analyzes your writing and tells you which famous writer you most resemble.”
So I have to try it. First try, a random blog post, produced this horror:
I write like
Isaac Asimov
Isaac Asimov
(Asimov: good ideas, bad writing.)
Second try, the first three pages from a paper I'm writing produced something even worse:
I write like
H. P. Lovecraft
H. P. Lovecraft
OK, it's a paper about ICANN, and thus something of a horror show, but even so…
Fortunately, Kevin Drum thinks there's nothing to it.
George Orwell’s “Why I Write” also returns H.P. Lovecraft. Go figure.
If it had said I write like Orwell, I would have been very flattered.
I just ran the first ten of H. P. Lovecraft stories from http://www.dagonbytes.com/thelibrary/lovecraft/ through it and got:
Like H. P. Lovecraft:
The Alchemist
The Book
The Call of Cthulhu
The Case of Charles Dexter Ward
Like George Orwell:
At the Mountains of Madness
Like James Joyce
Azathoth
Celephais
Like Mary Shelley
The Beast in the Cave
Beyond the Wall of Sleep
Like Neil Gaiman
The Cats of Ulthar
On this small sample, it seems to be about 40% accurate.
This kind of statistical analysis of texts made a stir in the 1980s when it was used for a case in, er, Sweden I think, to prove that some document had been written by a particular individual. I went to an AI seminar about it once: they had about 40 measures that they had discovered to be fairly constant, ranging from the density of the word “and” and the ratio of “a” to “an” (no kidding!). They tried a similar thing for music; I don’t know how much success they had, though.
If you’re interested in this kind of thing, check out http://www.philocomp.net/humanities/signature – it’s free for academic use.
Richard