I've been revisiting DATR, the lexical knowledge representation language, as a possible format for the next generation of MorphGNT. I was previously considering developing my own RDF/graph-based format but I suddenly remembered DATR from my student days and it makes a lot more sense to use it rather than try to build my own.
Looking at DATR material, I haven't seen anything more recent than 1998 so I'm not sure if it's still the state-of-the-art. It's a natural fit for some kind of RDFization, something I'm sure I'll eventually end up doing if someone hasn't already.
Of course, I'll have to write Python code to manipulate DATR. Again, unless some already exists. But I'm almost hoping not as I love implementing specs, especially using test-driven development.
UPDATE 2005-04-19: Now see DATR in Python