geometric analysis of web-documents

Hai all, 

For my final thesis I'm working on a project were I would like to
enrich the annotation of 'structured' webdocuments. Now I'am searching
for information on how to do a geometrical analysis of webdocuments.
All though the layout is desribed by the HTML i can't put my finger on
how to convert this information to a geometric annoted web-document.

Hope this message will return some new ideas, or other kind of
sollutions to this problem. Thx all lot for the effort.

Greetings,

Hans Klein Hesselink