Garbling Guardedly


The English language can be very precise with a lot of subtle choices of words to give that delicate nuance that the writer has being striving to work into his or her piece of prose while attempting to survive in a lonely garret on the crumbs from the rich publishers table. Meanwhile, hoards of scribblers are churning out impenetrable agreements, conditions of use, assembly instructions, instruction manuals and reports from the hydra-like bureaucracies from around the globe. Some of these works of art have been penned in other than English first and then been fuzzied-up further in translation.

I had better ‘fess up here. I have just taken liberties with the word “fuzzy.”  The temptation to get ahead of myself was too great. 

We are coming up to the first anniversary of the publication of “The Index of Garbledness,” IG, (1). You see the authors are all mathematicians with a skill in Fuzzy Set Theory. The aim of applying their algebraic manipulations was not to give a numerical value to the IG  of the instruction manual that came with your latest delivery of flatpack furniture. No, the ultimate aim was to crack codes.

Pairs of letters, with or without spaces make up the components of the fuzzy sets. The challenge was to select those fuzzy sets that are characteristic of English. These are called Feature Fuzzy Sets. Now 10 of these were chosen, and through the wonders of algebra, a series of texts were put to the test.

Spaces between words were dumped for the tests, as no self-respecting spy would help the opposition by separating the words. Three texts were used. The first had 10 to 50% of the words garbled, the second had 50 to 90% of the words garbled and the last was totally garbled. The IG values for the three cases were 9.75, 22.67 and 28.85.

Many of us who have had notes from offspring at college explaining why the cash you just sent was insufficient for their purposes will have a good feel for what an IG~10 means. The instructions for filling in most government forms can sometimes match that, but my guesstimate for the small print for loans is more like an 15<IG <20. How much longer, I wonder, before we can download an App from the Cloud to scan our Docs to give them an IG value? You developers out there – we need it and we need it now!

  1. P.K. Saxena, P. Yadav,. And G. Mishra, Defence Science Journal, 60, 415, (2010)

Leave a Reply