Simplified English is notoriously difficult to write — and not a lot of fun to read, either. Given a clean slate, how would a linguist go about creating a new and/or improved controlled language? We put this question to Lee Humphreys, formerly of the Eurotra UK group at the University of Exeter, more recently of SITE.
    Adopting the Simplified English approach to the software documentation domain, I would take a large corpus of texts accepted as high quality by technical authors and users alike. I would analyze the corpus — by hand — to find the constructions actually used. This list — edited if necessary — could form the basis of the grammatical specs for my simplified language. The problem, of course, is to find a reasonable way of presenting these specifications to the authors. The grammar you would specify for the parser is obviously not ideal for humans. And, er, yes, I would try to tighten up the general language lexical specifications as well.
    “ Humphreys and his colleagues at Exeter became interested in restricted language a number of years ago when word came out that Perkins Engines (Peterbor, ough, UK) achieved surprisingly good results feeding Controlled English to the Weidner English,to,French MT system.

(See  article  that this sidebar accompanied)