Generating compound nouns with the ERG

grammars ERG
1

I’m having issues with generating compound nouns with the ERG. I’m trying to generate “trash can,” but if I don’t specify number information I get stuff like “trashes can.” I tried to enforce the fact that the ARG0 of “trash” should be singular, but it won’t generate from an MRS with that information (ARG2 of compound shows the information, but I assume anywhere where x2 is found has that information too):

[ TOP: h19 
INDEX: e16 
RELS: < 
	[ unknown LBL: h18 ARG: x12 ARG0: e16 ]
	[ _the_q LBL: h15 ARG0: x12 RSTR: h13 BODY: h14 ]
	[ compound LBL: h11 ARG0: e8 ARG1: x12 ARG2: x2 [ x NUM: sg ] ]
	[ udef_q LBL: h7 ARG0: x2 RSTR: h5 BODY: h6 ]
	[ _trash_n_1 LBL: h3 ARG0: x2 ]
	[ _can_n_1 LBL: h11 ARG0: x12 ] > 
HCONS: < h5 qeq h3 h13 qeq h11 h19 qeq h18 > ]

When I parse “the trash can” with the ERG I get an MRS like this:

[ LTOP: h0 
INDEX: e2 [ e SF: prop ] 
RELS: < 
	[ unknown<0:13> LBL: h1 ARG0: e2 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ]  
	[ _the_q<0:3> LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ] 
	[ compound<4:13> LBL: h8 ARG0: e9 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: x4 ARG2: x10 ]  
	[ udef_q<4:9> LBL: h11 ARG0: x10 RSTR: h12 BODY: h13 ] 
	[ _trash_n_1<4:9> LBL: h14 ARG0: x10 ]  [ _can_n_1<10:13> LBL: h8 ARG0: x4 ] > 
HCONS: < h0 qeq h1 h6 qeq h8 h12 qeq h14 >
ICONS: < > ]

It seems to not record the fact that x10 is singular, but if I don’t enforce that on the generation side I end up with “trashes” can, so I’m not sure what’s going on.

Any help would be greatly appreciated!

2

The grammar includes a separate N-N compound rule to allow for less typical compounds that have a plural non-head, as in “systems analyst”. We don’t currently have a good way of blocking this rule in parsing for the majority of compounds where the non-head is not marked for number. But you can, if you wish, configure the ERG to exclude the unusual compound rule that allows the plural non-head: in the file erg/lkb/nogen-rules.set, add the following lines:

; Avoid plural non-head noun in N-N compounds
n-hdn_cpd-pl_c

Then recompile the grammar by moving to the erg directory and executing:
ace -G erg.dat -g ace/config.tdl

Now when you generate from an MRS that does not have a number marking for the non-head noun EP (the normal situation), you should get only the non-plural non-head noun. Note that you won’t be able to generate “systems analyst” with the grammar thus constrained.