Saturday, July 17, 2010
Bashing the romantic notion of the artist against the computational power of an algorithm and you get, if nothing else, amusing (and likely short-lived) internet memes. You may have heard of the “I write like” thing that a programmer in Montenegro, Dmitry Chestnykh, put together. Basically, you copy and paste some chunks of your or someone else’s prose into a window and it uses code developed for detecting spam to tell you which famous writer you “write like.” I write like Dan Brown, I was delighted to find. For fun, I had it analyze some Nigerian 419 spam emails and discovered that while most write like David Foster Wallace, “MISS STEPHANIE UJU” writes like Shakespeare. It’s received sufficient notoriety in the last few days to spark some media attention and even some backlash (originally, it would tell you which of thirty-seven white male authors and three white female authors you wrote like; apparently the canon has been opened up a bit in response).
Anyway, having randomly also just come across digital artist Jason Huff’s “AutoSummarize” project, however, an experiment presented itself. Huff took “the top 100 most downloaded copyright free books” and used Microsoft Word 2008’s AutoSummarize function to summarize them, in their entirety, into ten sentence versions (“Word has examined the document and picked the sentences most relevant to the main theme”). The result is sort of wonderful. Here, for example, is Mark Twain’s Huckleberry Finn:
Adventures of Huckleberry Finn
by Mark Twain
“All right. “All right. “Why, Jim?”
“Blamed if I would, Jim.”
“WHAT raft, Jim?”
And I’m satisfied. That makes me happy. But that gave me an idea: plug that in to the “Write like” program and see who it “writes like.” And guess what? It writes like Mark Twain!
Awesome. Like if Faulkner and Gertrude Stein were co-writing a book while drinking heavily. I like it (esp. since it was much shorter to read than Huck Finn. Or even any Faulkner or Stein!)
Your first two paragraphs write like Cory Doctorow
Well, I was upset to find I write like Dan Brown, until I found that another sample of my writing would be akin to Cory Doctorow.
Then again, the following is like Neil Gaiman:
No, not sausages. Not today.
No sausages today?
Well, not till sausage day.
So I’m not sure, without his statistical methods being made available, that there’s much weight we should put on it