Can it spot a cliché or catch a redundant phrase? Yes. Can it evaluate whether your syntax and diction are in line with other books in your genre? I think so. Can it predict whether your novel will be a bestseller or evaluate your story arc? Not yet.
Authors.me, in conjunction with Publisher Weekly’s BookLife, recently launched what they call an Intelligent Editorial Analysis Report, which uses machine learning to compare manuscripts to bestsellers “across hundreds of literary, syntactical, and plot-based metrics.” Curious to see how it worked, I sent in the manuscript of a novel I recently edited.
The report I got back, similar to the samples you can see here, had some promising insights as well as some obvious flaws.
SHOW ME THE DATA
The Intelligent Editorial Analysis was able to spot clichés and redundant phrases with reasonable accuracy. The report also provided a helpful breakdown of the manuscript by parts of speech, revealing that the author may be relying too heavily on adverbs. The word length analysis would be useful for authors and editors writing for middle grade or young readers.
The drawback? The report provides a scattering of examples rather than a complete list of problems. Getting back a list of every cliché, redundant phrase, mixed metaphor, and adverb would be quite useful. Instead, the report showed me one redundant phrase from the manuscript I uploaded and then linked to a list of common redundancies. I’d also like to see a more complex graph of the story arc (wrongly identified as a riches-to-rags plot), like the ones Authors.me shows for Gone Girl and 50 Shades of Grey in this blog post.
What did it miss? Among the things I expected it to catch, Author.me’s analysis didn’t flag comma splices or misused semicolons. While not perfect, machines can do a reasonably good job of catching such straightforward problems. And then there is the long list of problems I didn’t expect a machine to catch, like the minor character who gets three pages of dialogue in chapter four and then disappears, never to be seen again.
FIX THE COMPS
The biggest flaw in Author.me’s story analyzer may be their comps. The report gave the novel a score of 87 out of 100 for commercial viability. Without any other data points, I’d agree with that score. The novel has a compelling main character and a suspenseful plot, and I expect it to do well in its category. But what does the Intelligent Editorial Analysis compare this novel to? We don’t know, beyond a general statement that the software has “analyzed thousands of bestsellers.”
The syntax and diction section of the report does provide a dozen specific comps, none of which are on target. The manuscript is a fantasy novel for adult readers with a main character who is a witch. Useful comps would be novels like Alice Hoffman’s Practical Magic and Deborah Harkness’s A Discovery of Witches. Instead, the Authors.me report compares it to a handful of Harry Potter novels, plus a half-dozen classic novels like Edith Wharton’s Age of Innocence (which is a caustic portrait of Gilded Age society, not a novel about infants as the baby rattle icon illustrating it would lead you to believe.)
WHAT ELSE CAN MACHINE ANALYSIS TELL US?
I’m a believer in the power of machines to unearth buried patterns or identify details that can escape human readers. I’d love to see Authors.me use their tool to give authors and editors insights into the kinds of questions that we now use experience and instinct to answer. For example, what percentage of bestsellers, in which genres, are written in present rather than past tense? Is this trend still on the rise, or is it beginning to fall off? How many sex scenes are in bestselling contemporary romances, and where are they placed? I can think of a dozen questions I believe machines could help us answer.
The Intelligent Editorial Analysis has the potential to be a tool which helps authors and editors spend more time thinking about the how of revisions, rather than the what or the why. It’s worth keeping an eye on but not worth your $100 (or even the $50 intro price) . . . yet.