TODO:

fix lexicon builder to throw out duplicate lexemes

record ersatzing
improved MRS extractor

make ambiguity packing work with semantics (and record packing)
statistical parse selection?

OPTIMIZATIONS:

full subsumption packing

hyperactive excursion



(considered)
+ pick cleverly between dg1 and dg2 as target in unify1
+ optimize lexicon storage -- probably get down by several times in size
+ key vs. head driven = "KEY-ARG" inside rule daughters
- smarter ordering of agenda