A while back I had written a new documentation system called unidoc, this technology is different then most other systems out there because it is a language independent. Rather then trying to decypher what is going on within a bit of code the parser leaves that up to the developer (which is what most documentation systems do anyway). Documentation of what various bits of code are doing is first defined through a very rudamentary language.
The parser extracts out the comments and generates a xml data stream sending it to a langauge-specific xslt template, some data items are parsed in such a way that the xslt can further process, define, and lay out langauge specific properties. In the end generating language specific documentation through a xslt definition. Right now things are pretty simple, it took a weekend to write the whole thing but what I would like to do is re-write it using antlr+xslt, because right now things are just a simple regex parser that I wrote.
I also have a planned extension to the processor which focuses more on outputting a data format instead of documentation, and finally I think it would be neat to wrap this all up with a publishing system of some sort effectivly allowing people to comment on documentation.
I would also like to create a few simple xslt templates as proof of concept, so if anyone is interested in any of what I am talking about drop me a line and I will have something for you to do.