lack of code documentation
I'd be really curious to see if the output is comprehensible and informative. There's a huge training corpus for this do I'd assume this would work.
1. document should be about why, not what.
2. not all decisions can be revealed in code itself. (for example, adding a delay to workaround a hardware bug.)
3. the biggest part missing from code documentation is the big picture document, not really the code level document. even you know what a class/function is, you don't know how it is used in the entire project, its life cycle, etc.