Negated relative clause used for parsing not appearing when generating

grammars ERG
1

I found another case where the string I’m using for parsing isn’t appearing when I use the same MRS for generation.

The string I parsed was the student who does not bake a cookies and I get this result:

[ TOP: h0
  INDEX: e2 [ e SF: prop ]
  RELS: < [ unknown<0:37> LBL: h1 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ARG0: e2 ]
          [ _the_q<0:3> LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ]
          [ _student_n_of<4:11> LBL: h8 ARG0: x4 ARG1: i9 ]
          [ neg<21:24> LBL: h8 ARG0: e10 [ e SF: prop TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: h11 ]
          [ _bake_v_cause<25:29> LBL: h12 ARG0: e13 [ e SF: prop TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: x4 ARG2: x14 [ x PERS: 3 NUM: pl IND: + ] ]
          [ udef_q<30:37> LBL: h15 ARG0: x14 RSTR: h16 BODY: h17 ]
          [ _cookie_n_1<30:37> LBL: h18 ARG0: x14 ] >
  HCONS: < h0 qeq h1 h6 qeq h8 h11 qeq h12 h16 qeq h18 > ]

But when generating I get these results:

The student cookies are not baked by
The student cookies are not baked by.
The student cookies are not baked by
The student cookies are not baked by.
The student cookies are non baked by
The student cookies are non baked by.

What’s happening here?

2

I made a typo in the original post, the string was the student who does not bake cookies

3

Another update on this.

It seems that the results change in an unexpected way when I generate from an MRS with a relative clause vs. a negated version of that MRS.

Original MRS:

[ TOP: h0
  INDEX: e2 [ e SF: prop ]
  RELS: < [ unknown LBL: h1 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ARG0: e2 ]
          [ _the_q LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ]
          [ _student_n_of LBL: h8 ARG0: x4 ARG1: i9 ]
          [ _bake_v_cause LBL: h8 ARG0: e13 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] ARG1: x4 ARG2: x14 [ x PERS: 3 NUM: sg IND: + ] ]
          [ _a_q LBL: h15 ARG0: x14 RSTR: h16 BODY: h17 ]
          [ _cookie_n_1 LBL: h18 ARG0: x14 ] >
  HCONS: < h0 qeq h1 h6 qeq h8 h16 qeq h18 > ]

Generation results:

The student by which a cookie was baked.
The student by who a cookie was baked.
The student by which a cookie was baked
The student by who a cookie was baked
The student that a cookie was baked by
The student a cookie was baked by
The student that a cookie was baked by.
The student, by which a cookie was baked.
The student, by who a cookie was baked.
The student which a cookie was baked by
The student, which baked a cookie.
The student who a cookie was baked by
The student a cookie was baked by.
The student which a cookie was baked by.
The student who baked a cookie
The student, which a cookie was baked by.
The student who a cookie was baked by.
The student who baked a cookie.
The student that baked a cookie
The student that baked a cookie.
The student, who baked a cookie.
The student, that a cookie was baked by.
The student which baked a cookie
The student which baked a cookie.
The student, who a cookie was baked by.
The student, that baked a cookie.

Negated MRS:

[ TOP: h0
  INDEX: e2 [ e SF: prop ]
  RELS: < [ unknown<0:34> LBL: h1 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ARG0: e2 ]
          [ _the_q LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ]
          [ _student_n_of LBL: h8 ARG0: x4 ARG1: i9 ]
          [ neg LBL: h8 ARG0: e10 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] ARG1: h11 ]
          [ _bake_v_cause LBL: h12 ARG0: e13 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] ARG1: x4 ARG2: x14 [ x PERS: 3 NUM: sg IND: + ] ]
          [ _a_q LBL: h15 ARG0: x14 RSTR: h16 BODY: h17 ]
          [ _cookie_n_1 LBL: h18 ARG0: x14 ] >
  HCONS: < h0 qeq h1 h6 qeq h8 h11 qeq h12 h16 qeq h18 > ]

Generation results:

The student a cookie was not baked by
The student a cookie was not baked by.
The student a cookie was not baked by
The student a cookie was not baked by.
The student a cookie was non baked by
The student a cookie was non baked by.

Questions:

  1. Why does the original version have the complementizer versions only, but not the ones without it?
  2. Why does the negated version only have the ones without the complementizer?

I feel like I expected both MRSs to give me both kinds of results.

4

This unexpected behavior is due to the same bug in the grammar that was reported in the thread “ERG not generating *any* outputs for same MRS received from parsing”, due to a missing trigger rule that would introduce relative pronouns into the generator chart for negated relative clauses. As noted in that thread, the bug is fixed, and the correction will be included in the next release.