Hi Pander, I have the details not ready (I can look them up, if you don't know them already), but at the EuroTeX conference in Arnhem 1995 there was a talk about hyphenation patterns for compounds generated with patgen (using a correctly hyphenated list of compound words as input). The author claimed good success for finding correct Haupttrennstellen (of course, some words linke german Staubecken or Wachstube are unhyphenatable with this approach).
Keep me in touch with your work, it is very interesting. Unfortunately, E-TeX never implemented different weights for hyphenation points. --Jörg Knappen P.S. Another (very different) approach was the Austrian SiSiSi project, still to be found somewhere on CTAN (in the form of documented change files to TeX.web!!!)
