Problems yet to solve

Matching issues - context, multiple parts, multiple matches There are plenty of potential cases where a regular expression won't return the expected output, especially if users are perverse. There are ways to create more robust regular expressions, of course. Also, there may be times when certain matches are appropriate only in particular contexts. I haven't yet begun to address this, but may take a look at XPath.
Matching on markup AND/OR lexical content There may be times when the processor should test everything matching particular rules and apply transformation or annotation. I haven't set this up yet.