Generating unknown words

Is there a way to generate with the ERG from an MRS that has an unknown word?

For example, I can parse “the stumpy tree” but “stumpy” is unknown:

[ TOP: h0
  INDEX: e2 [ e SF: prop ]
  RELS: < [ unknown<0:15> LBL: h1 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ARG0: e2 ]
          [ _the_q<0:3> LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ]
          [ _stumpy/jj_u_unknown<4:10> LBL: h8 ARG0: e9 [ e SF: prop TENSE: untensed MOOD: indicative PROG: bool PERF: - ] ARG1: x4 ]
          [ _tree_n_of<11:15> LBL: h8 ARG0: x4 ARG1: i10 ] >
  HCONS: < h0 qeq h1 h6 qeq h8 > ]

Would adding something to the SEMI be enough?

We have never been able to devise a method enabling the ERG to generate from unknown semantic predicates (apart from unknown proper names, where the predicate is known and only the CARG attribute varies). The lexicon is currently being expanded significantly to include all words in WordNet (about 140,000 stems), which should reduce the frequency of running into a missing semantic predicate when generating, but that enhancement won’t help you in the near term.

As a hack, you can add a lexical entry with the right lexical type to erg/lexicon.tdl, and add the corresponding line(s) to erg/etc/surface.smi, then recompile. For example, for the adjective “stumpy” here is the entry to add to lexicon.tdl:

stumpy_a1 := aj_-_i_le &
[ ORTH < “stumpy” >,
SYNSEM [ LKEYS.KEYREL.PRED “_stumpy_a_1_rel”,
PHON.ONSET con ] ].

And here is the corresponding line to add to erg/etc/surface.smi:

_stumpy_a_1 : ARG0 e, ARG1 x.