Hello, Don’t Read faithful!
Quick post because I just found out about Booklamp and the Book Genome project.
The Book Genome project uses computer algorithms to analyze a book’s content and style based on a large number of variables and it sounds very legit. The data are then aggregated on Booklamp.org and used to provide similar title recommendations.
It sounds like a great idea, except I feel it has a fundamental flaw. The analysis of language part I love and am down for. But the analysis of subject matter is what sticks in my craw.
I don’t read one genre. I don’t read one author. I read GOOD, compelling writing. And whether or not a book has the same amount of cigarette smoking is not an issue for me in selecting fiction. I shit you not, cigarette smoking is a dimension by which the books on Booklamp are measured. As well as partying/deviance and scheduling/elements of time. Again, not things I find important in fiction.
Case in point: I hate Vampires. But I liked Dracula and I LOVE Let The Right One In–because they have great characters and plot. Interview With a Vampire? HATED IT! Because it was boring and self-important and lacked a good plot.
Am I saying Booklamp is bad? Far from it. I think it’s simply misguided. Because at the end of the day, it’s gonna be another tool Amazon.com can use to say, “Hey kids! If you liked X you’ll love Y!” in those lovely ads they show everywhere.
But that’s not something a computer can pick up on.
I was reading about their variables and I saw not one thing that measures plot strength. They do have some good tools to measure language, but it says nothing about how well the events of the story are tied together through cause and effect relationships(i.e., Plot). Can this be measured with an algorithm?
Can te Book Genome Project grasp the subtlety of language involved in metaphor? Can it detect the thematic similarities between an allegorical story and a literal one? How about breaches of the fourth wall? Can it recognize them and count them?
I’d love to see how Book Genome analyzes a cut-and-paste patchwork of disparately themed and dense literature. Could it detect that there was no actual plot to the story? Would it pick up on the fact that there is no discernible climax, or denouement?
To be fair, they adress their shortcomings in an FAQ.
I’m probably sounding like a crotchety-technophobe grandpa here, but I know just because a computer can’t do this doesn’t mean people shouldn’t try. I just wish they used some better variables and put it to a grander use–like literary analysis–than to semi-scientific recommendations.
What do you think? Book Genome: Waste of Time or the future of Literacy?
Leave a comment.