With our treebanks, it's fairly easy to generate constructions with various constraints (e.g. [1]) and my "New Kind of Graded Reader" algorithms from 2008 / 2010 provide somewhat of a way of ordering them for ability to read more, faster and even inline them in English for context[2]. The problem at the time, was of course, the morphological stuff which is why I went back to morphology the last few years, to get that in better shape
![Smile :-)](./images/smilies/icon_e_smile.gif)
James
[1] http://jktauber.com/2010/04/14/all-subt ... t-clauses/
[2] http://jktauber.com/2010/04/25/inline-r ... nt-john-2/