Bad MRS for "Abrams wondered which dog barked"
goodmami opened this issue · 1 comments
goodmami commented
Item 291 in the mrs
test suite gives the following bad MRS in the 2018 version and trunk:
[ "Abrams wondered which dog barked."
TOP: h0
INDEX: e2 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ]
RELS: < [ proper_q<0:6> LBL: h4 ARG0: x3 [ x PERS: 3 NUM: sg IND: + ] RSTR: h5 BODY: h6 ]
[ named<0:6> LBL: h7 ARG0: x3 CARG: "Abrams" ]
[ relative_mod<7:33> LBL: h1 ARG0: i9 ARG1: x10 [ x PERS: 3 NUM: sg IND: + ] ARG2: h11 ]
[ udef_q<7:33> LBL: h12 ARG0: x10 RSTR: h13 BODY: h14 ]
[ generic_entity<7:33> LBL: h15 ARG0: x10 ]
[ _wonder_v_1<7:15> LBL: h1 ARG0: e2 ARG1: x3 ]
[ _which_q<16:21> LBL: h16 ARG0: x10 RSTR: h17 BODY: h18 ]
[ _dog_n_1<22:25> LBL: h19 ARG0: x10 ]
[ _bark_v_1<26:33> LBL: h11 ARG0: e20 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] ARG1: x10 ] >
HCONS: < h0 qeq h1 h5 qeq h7 h13 qeq h15 h17 qeq h19 > ]
Note that udef_q
, generic_entity
, _which_q
, and _dog_n_1
all share the ARG0
of x10
.
This analysis is not present in the 1214 version, so it appears to be a regression.
danflick commented
The bad MRS came from an unwanted parse, which the ERG now excludes.