Thanks for the link. Do you have any info on the performance of this parser?
I welcome the move away from regular expressions though - they are just not necessary in this particular domain. We'll see if PatternDB's coarse grained approach comes back to bite me.
I'm happy to help as I can if you decide to use PatternDB - you can find me at l.skibinski at elifesciences dot org. I have some notes for getting started quickly I really should publish ...