Just got fed up with reading a convoluted government website and went looking for a fog index algorithm to feed it to. Found a nice one here. The financial regulation that is still boggling me got off with an index of 16.50, which is surprisingly low, and a "Flesch Reading Ease" of 34.47 on a 100-point scale, which seems about right.
My question, though, is whether mechanically computing fog indices performs any useful service. Can a machine decide if a phrase is confusing or not? What about, for example, very simple phrases that cause confusion by oversimplifying, as in the school textbook that mystified Richard Feynman with the meaningless phrase, "Energy makes it go"? What about IRS or Social Security publications written in such "plain language" that they spackle over important technicalities?
How complicated is just complicated enough?
Posted by Martha Bridegam at January 9, 2005 10:50 PMAccording to the fog algorithm Horizon itself ranks at about the level of "Most popular novels".
The Flesch scale comes weighs in at 71.24, though. Quite a discrepancy.
On the other hand, my old zine Vox Clamantis comes in at "academic paper" level for fog but a mere 35 on the Flesch scale.
Anyway, these scales are nothing. Take a look at some of the stuff they've been designing lately:
CyberProf: an Intelligent Human-Computer Interface for Asynchronous Widearea Training and Teaching
This one uses physical number theory, fuzzy logic*, and so on to grade student answers.
Automatic Essay Grading Using Text Categorization Techniques This one has a little overview of attempts to write software that grades undergraduate essays. Basically, counting the number of vocabulary words used that are considered topical to the question has been found to be surprisingly accurate, though it looks like this one is more sophisticated than that and uses neural networks.
So it appears that you can't knock algorithms in general for not getting context or being expert in the topic at hand. There are probably similar shortcuts that could be used, if anyone cared enough, to give a meaningful complexity score.
* I've read recently that at some point in the '70s, some Generative Semanticists briefly posited something called "Fuzzy Grammar". This has been interpreted by Chomskyites (or at least the Chomskyite who wrote the article I read) as one of the final death throes of Generative Semantics, a theory which competed unsuccessfully with Generative Grammar. It seems that whenever a field gratuitously borrows a term from another (cf., "Generative Anthropology"), the borrowing always goes down the scientific ladder, and the term is usually applied to a theory that is far less plausible than is the standard even in its own field. Is this the law of the conservation of plausibility at work?
Posted by: Alan Hogue at January 10, 2005 11:24 AMSince it's inception, Horizon clocks in at a 12th grade reading level with a Flesch scale at 62.79. However, January to date, it's operating at an 11th grade level.
The following two sentences register at college junior with a Flesch of 47.60:
The mechanics are driven by average sentence length and average syllables per word. Short sentences with simple words translate into greater comprehension on the part of the reader.
Rewritten, they come in at a 9th grade level and a Flesch of 81.86:
The formula is based on sentence and word length. A reader is more likely to understand writing that is brief and to the point.
I've been playing with this thing all day and I think it says more about personal writing styles than actual comprehensibility (7 syllables - big score).
Posted by: Bobby Farouk at January 10, 2005 12:51 PMAccording to the fog algorithm Horizon itself ranks at about the level of "Most popular novels".
I think we should take that as a compliment.
Posted by: Alan Allport at January 11, 2005 12:30 PMI don't know if I'd go so far as to call it a compliment, but it is a good thing.
Still, going to the kind of high school I went to, I can't help but wince at the idea of writing anything at a truly 12th grade reading level. I much prefer the appellation "modern popular novel". Cosmetic, I suppose, but important.
Posted by: Alan Hogue at January 12, 2005 09:38 AMBurmese Days (Chapter 16):
Grade Level - 6th
Flesch Score - 82.26
The Lion&the Unicorn (Part 1):
Grade Level - College sophomore
Flesch Score - 58.61
Why I Write:
Grade Level - 11th
Flesch Score - 67.53