'Knock on the door" using 2020 grammar: shouldn't there be an MRS for "[you] knock on the door [and I don't care where you are]"?

Below are the 3 readings I get from the phrase “knock on the door” using ACE. Shouldn’t I also be getting a reading that has a 3rd person pronoun (pronoun_q + pron) and the locative preposition _on_p_loc? I.e. the interpretation being something like:

"[you] knock on the door [and I don't care where you are]"

Am I just missing something?

% ace -g erg-2020-osx-perplexity.dat -Tf

knock on the door

SENT: knock on the door
[ LTOP: h0
INDEX: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ]
RELS: < [ pronoun_q<0:17> LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron<0:17> LBL: h7 ARG0: x3 ]
[ _knock_v_1<0:5> LBL: h1 ARG0: e2 ARG1: x3 ARG2: i8 ]
[ _on_p_state<6:8> LBL: h1 ARG0: e9 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: e2 ARG2: x10 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<9:12> LBL: h11 ARG0: x10 RSTR: h12 BODY: h13 ]
[ _door_n_1<13:17> LBL: h14 ARG0: x10 ] >
HCONS: < h0 qeq h1 h5 qeq h7 h12 qeq h14 >
ICONS: < > ]
[ LTOP: h0
INDEX: e2 [ e SF: prop-or-ques ]
RELS: < [ unknown<0:17> LBL: h1 ARG0: e2 ARG: x4 [ x PERS: 3 NUM: sg IND: + ] ]
[ udef_q<0:17> LBL: h5 ARG0: x4 RSTR: h6 BODY: h7 ]
[ _knock_n_1<0:5> LBL: h8 ARG0: x4 ]
[ _on_p_loc<6:8> LBL: h8 ARG0: e9 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: x4 ARG2: x10 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<9:12> LBL: h11 ARG0: x10 RSTR: h12 BODY: h13 ]
[ _door_n_1<13:17> LBL: h14 ARG0: x10 ] >
HCONS: < h0 qeq h1 h6 qeq h8 h12 qeq h14 >
ICONS: < > ]
[ LTOP: h0
INDEX: e2 [ e SF: prop-or-ques TENSE: tensed MOOD: indicative PROG: - PERF: - ]
RELS: < [ _knock_v_1<0:5> LBL: h1 ARG0: e2 ARG1: i3 ARG2: i4 ]
[ _on_p_state<6:8> LBL: h1 ARG0: e5 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: e2 ARG2: x6 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<9:12> LBL: h7 ARG0: x6 RSTR: h8 BODY: h9 ]
[ _door_n_1<13:17> LBL: h10 ARG0: x6 ] >
HCONS: < h0 qeq h1 h8 qeq h10 >
ICONS: < > ]

Why 3rd person? “you” is 2nd person.

And _on_p_state is locative. It’s a subtype of _on_p_loc, although this is an easy point to overlook; the predicate hierarchy is here: svn.emmtee.net/erg/tags/2020/etc/hierarchy.smi

The first parse (the non-fragment one) looks correct to me. As Guy says, the implicit pronoun “you” supplied for imperatives is second-person, as shown in the MRS, not third-person (which is for “he/she/it/they”). And to give a little more detail on locative prepositions, the ERG distinguishes directional from stative locatives, with the contrast illustrated in a sentence such as “the mouse ran under the table” where the stative reading has the mouse staying under the table and running around there, while the directional reading has the mouse moving from some other spot in the room on a path that takes it under the table. The lexical entry for “on” has a predicate “_on_p_loc” which can be specialized during parsing to either “_on_p_state” or “_on_p_dir”. In the MRS for the first parse of your sentence, you see that “_on_p_state” is used, and the external argument of that predicate (ARG1, the thing being located) is the event (ARG0) of _knock_v1. So as you expected, the MRS says nothing about the location of “you”, but declares that the knocking happened on the door.

@guyemerson: Sorry, yes I meant 2nd person. And you’re right that I didn’t realize _on_p_state was a subtype of _on_p_loc. And thanks for the summary of the locative stative or directional readings, @Dan.

For some reason I got it stuck in my head that stative prepositions were always about the state of the actor. Probably due to the scenarios I was exploring when I first encountered them. I wasn’t (but should be) paying attention to the ARG1/ARG2. I get it now.

I’m still confused about what _on_p_loc means when used directly vs. its stative or directional subtypes. Does it have a different meaning, or is it just implying “underspecification” between the stative and directional subtypes?

For example, the first two readings of ‘Look on the table’ are below. The first one uses _on_p_loc and the second _on_p_state. Both of these _on_p alternatives have the same ARG1 and ARG2 which (effectively) point at _look_v_1 and table_n_1 respectively and everything else in the MRS is the same.

What are the different meanings of the two?

'Look on the table.'

[ LTOP: h0
INDEX: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ]
RELS: < [ pronoun_q<0:18> LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron<0:18> LBL: h7 ARG0: x3 ]
[ _look_v_1<0:4> LBL: h1 ARG0: e2 ARG1: x3 ]
[ _on_p_loc<5:7> LBL: h1 ARG0: e8 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: e2 ARG2: x9 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<8:11> LBL: h10 ARG0: x9 RSTR: h11 BODY: h12 ]
[ _table_n_1<12:17> LBL: h13 ARG0: x9 ] >
HCONS: < h0 qeq h1 h5 qeq h7 h11 qeq h13 >
ICONS: < > ]

[ LTOP: h0
INDEX: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ]
RELS: < [ pronoun_q<0:18> LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron<0:18> LBL: h7 ARG0: x3 ]
[ _look_v_1<0:4> LBL: h1 ARG0: e2 ARG1: x3 ]
[ _on_p_state<5:7> LBL: h1 ARG0: e8 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: e2 ARG2: x9 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<8:11> LBL: h10 ARG0: x9 RSTR: h11 BODY: h12 ]
[ _table_n_1<12:17> LBL: h13 ARG0: x9 ] >
HCONS: < h0 qeq h1 h5 qeq h7 h11 qeq h13 >
ICONS: < > ]

Yes. So the parse with _on_p_state has a more specific meaning than the parse with _on_p_loc (since everything else is the same).

1 Like

Ah, but you uncovered a bug in the grammar, since the underspecified _on_p_loc should normally be specialized either to _on_p_dir or _on_p_state. This expected specialization was failing for the small set of prepositions “in, on, at” which can also have a temporal sense. The fix will be included in the next release.

2 Likes

Good to know! So, once the bug is fixed, does that mean _on_p_loc shouldn’t ever show up on its own?

For modification of verbal and adjectival phrases, yes, I think you should not expect to see _on_p_loc in an MRS for a parsed sentence. But you will still see this underspecified predicate for PPs modifying nouns, since the grammar does not currently try to draw a locative vs. stative distinction for modifiers of nouns. That is, for a sentence such as “any movement under the table will be detected” the resulting MRS just supplies the underspecified _under_p predicate. It’s clear that such modifiers are not always stative, since overtly directional PPs can modify nouns: “a sudden shift of weight onto the back wheels will flatten them”. For now, the grammar doesn’t try to be more specific, lacking a good theory of what would be correct.

2 Likes