This is a difficult question. Many trying-to-be-writers will leap onto a particular favourite author and say “I write like them!”. Once they’ve gotten a hundred thousand words down in various projects, I see these writers hesitate to be identified with a particular writer.
We all write differently. It’s a stupidly obvious statement, of course, because we are all different people. Different life experiences, different educations, different value systems. Yet we still want to write like a favourite writer.
There was a website that did the rounds amongst my friends the other day called “I Write Like“. You paste some of your writing, it does some word-analysis and tells you who it thinks you write like. We were having fun posting different fragments of our work to see what results we got. One friend consistently got H. P. Lovecraft and Stephen King. Not unusually, she writes dystopian fiction. I get authors from all over the shop: J. K. Rowling, Edgar Allen Poe, Stephen King, plus others I’d never really heard of. That night I got Neil Gaiman twice from two different pieces of writing.
There is something alluring about a computer program that can sum of your writing with another author’s name. The sense of validation when it is someone you like and value is very nice. And then there is the sense of horror when it is someone widely regarded as terrible, such as Stephanie Meyer. But I always had a suspicion we were fooling ourselves with this. I know Terry Pratchett writes quite differently now than twenty years ago. This is easy to see if you read an early Discworld novel followed by a later one. And if you read them all in publication order, you can see a progression in his writing ability and style. How could a computer program capture all that?
Turns out the author of “I Write Like” has made the source code available. Unfortunately for me, it is in a language I don’t know that uses paradigms I don’t like. So I have to take the author’s word that it uses a simple Bayesian filter. This makes sense, really, as that’s the algorithm the best spam filters use to learn what is and isn’t spam.
I’m kind of disappointed. Textual analysis can be very interesting and you could get a lot of metrics from a passage, if you really wanted to. Not just word frequencies, but things like sentence lengths and paragraph sizes over time, patterns of similar word use, frequency of dialogue use, etc, etc. Comparing them to existing authors’ works would be hard, though. At best, you’d get a range of percentages of how much you’re like other authors in different ways.
Whilst it could be fun to do, I don’t really want to spend the time doing it. I’d much rather have real people read my writing and, if necessary, compare it to other authors. If a simplistic computer program has trouble pigeon-holing me, I think that’s good.