Tip Tuesday #90

Tip Tuesday is a recurring feature where blog readers send in tips for fellow writers. If you'd like to send in a tip, please e-mail me at agentspotlight(at)gmail(dot)com.

Today's tip was sent in by Deren Hansen at the Laws of Making. It's a good one, like all Deren's tips! Make sure to stop by his blog when you're done reading here.

One of the persistent fantasies of the computer age is that if we give the machine the right data it will answer all our questions. From dating services that will find our perfect match to product recommendation systems that keep us abreast of the latest thing I-didn't-know-existed-but-now-desperately-need, computers are our new oracles.

Wouldn't it be nice if the machine could tell us whether our manuscripts were any good?

Perhaps, but computers don't work that way (something for which we should be grateful because if they did publishers would forget about authors and hire engineers to program an endless stream of bestsellers). What they can do is count, classify, and run statistics. None of that will answer an ultimate question like, "is my manuscript good?" but it can give you additional insights into your text.

For example, the Gender Genie, (mentioned in Tip #66 by Laura Lascarso) uses word frequency statistics to infer the gender of the author. If your main character doesn't share your gender and the genie reports your writing matches your gender more strongly than your character's gender, you may want to revise. If, on the other hand, the scores are relatively close, then that statistic doesn't give you much guidance.

There are a great many other text metrics, from word and phrase analysis (distribution, frequency, length) to readability and grade level. Among the web-based offerings, two that I like are Textalyser.net and UsingEnglish.com.

Textalyser is good for a quick check: enter sample text and study the analysis. It's a very simple site with interleaved advertising. You can use it anonymously.

UsingEnglish.com is targeted at teachers and ESL students, and requires registration (free). Once registered, you can store and analyze up to 20 documents. It's ideal if you want to compare sample text from throughout your manuscript to see, for example, if readability or grade level drifts. Of course you can delete text from the site once you no longer need it.

While I don't think you're in danger of anyone misappropriating your manuscript at either site, I suggest that you analyze text samples of a few thousand words. Analyzing your entire manuscript will only give you overall measures. Analyzing sample text every few chapters will give you a sense of trends and trouble spots.

I've posted a few more thoughts about text metrics on my blog, The Laws of Making.

Deren Hansen