"open all of the drawers" and generic_entity

Hi all! If I parse “Open all of the drawers” using erg-2018-osx-0.9.31 I get parses which fall into two main groups and I want to make sure I’m understanding them.

The first is this one, which I believe should be interpreted as “Open all of the different parts of the drawers”:

[ TOP: h0
INDEX: e2
RELS: < 
[ _the_q LBL: h14 ARG0: x13 [ x PERS: 3 NUM: pl IND: + ] RSTR: h15 BODY: h16 ]
[ _drawer_n_of LBL: h17 ARG0: x13 [ x PERS: 3 NUM: pl IND: + ] ARG1: i18 ]
[ pronoun_q LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron LBL: h7 ARG0: x3 [ x PERS: 2 PT: zero ] ]
[ _all_q LBL: h9 ARG0: x8 [ x PERS: 3 NUM: pl ] RSTR: h10 BODY: h11 ]
[ part_of LBL: h12 ARG0: x8 [ x PERS: 3 NUM: pl ] ARG1: x13 ]
[ _open_v_cause LBL: h1 ARG0: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: x3 ARG2: x8 ]
>
HCONS: < h0 qeq h1 h5 qeq h7 h10 qeq h12 h15 qeq h17 > ]

The second is this one, which I think is meant to mean “Open each and every drawer”

[ TOP: h0
INDEX: e2
RELS: < 
[ _the_q LBL: h15 ARG0: x14 [ x PERS: 3 NUM: pl IND: + ] RSTR: h16 BODY: h17 ]
[ _drawer_n_of LBL: h18 ARG0: x14 [ x PERS: 3 NUM: pl IND: + ] ARG1: i19 ]
[ pronoun_q LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron LBL: h7 ARG0: x3 [ x PERS: 2 PT: zero ] ]
[ _all_q LBL: h9 ARG0: x8 [ x PERS: 3 ] RSTR: h10 BODY: h11 ]
[ generic_entity LBL: h12 ARG0: x8 [ x PERS: 3 ] ]
[ _of_p LBL: h12 ARG0: e13 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: x8 ARG2: x14 ]
[ _open_v_cause LBL: h1 ARG0: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: x3 ARG2: x8 ]
>
HCONS: < h0 qeq h1 h5 qeq h7 h10 qeq h12 h16 qeq h18 > ]

First: do I have that right?

Second: Am I right in thinking that generic_entity is there in the second parse as a way to indicate that this interpretation is interpreting “all of the drawers” in a way that doesn’t have a “left side” of the “of the drawers” (as is described here)?

With the 2018 ERG you are using, I get two distinct analyses (with a mysterious duplicate parse for each that I’ll investigate, but ignore here). The one with the part_of' relation is not intended to mean "all of the parts of each drawer", but instead "every individual in the set denoted by the drawers’". (Note that we use this same predicate part_of' for "all of the rice" where "rice" can't be individuated like "the drawers" can, so here the intended interpretation is roughly "the complete content of the entity denoted by the rice’". ) In the second analysis, which introduces the generic_entity' and possessive of_p’ relations, the intended (and much less likely) meaning is “all of the implicitly identified individuals or content possessed by `the drawers’”. A clearer example where this second reading is the intended one is the following: “We took two samples of the varnish on the desktop, three samples of the sides, and none of the drawers” (meaning no samples of the varnish on the drawers). It’s harder to give a good example with “all of the drawers” instead of “some of” or “none of”, but consider “We don’t have enough photos of Kim for the yearbook, but we have all that we need of Sandy” where there is an additional restrictor overtly supplied for the implicit set denoted by “all”.

1 Like

Thanks, Dan! I think I understand the generic_entity reading now, but I’m still missing something for the part_of reading:

If I interpret the part_of(x1, x2) predication to mean: “x1 is an individual in the set denoted by x2”, then how do I interpret the below reading of “what is part of the cabinet?” It would only ever set x1 to be the actual individual that represents the cabinet, right? Which predication is doing the work of breaking the cabinet into parts?

[ TOP: h0
INDEX: e2
RELS: < 
[ _part_q LBL: h11 ARG0: x3 [ x PERS: 3 NUM: sg ] RSTR: h12 BODY: h13 ]
[ _the_q LBL: h14 ARG0: x10 [ x PERS: 3 NUM: sg IND: + ] RSTR: h15 BODY: h16 ]
[ _cabinet_n_1 LBL: h17 ARG0: x10 [ x PERS: 3 NUM: sg IND: + ] ]
[ part_of LBL: h9 ARG0: x3 [ x PERS: 3 NUM: sg ] ARG1: x10 ]
[ which_q LBL: h6 ARG0: x5 [ x ] RSTR: h7 BODY: h8 ]
[ thing LBL: h4 ARG0: x5 [ x ] ]
[ _be_v_id LBL: h1 ARG0: e2 [ e SF: ques TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: x3 ARG2: x5 ]
>
HCONS: < h0 qeq h1 h7 qeq h4 h12 qeq h9 h15 qeq h17 > ]

This is done by the model, not by a predication. (In the model-theoretic sense of “model”.)

Consider a similar example: “Who is part of the cabinet?” In a particular discourse context, we might know which cabinet is being referred to, and look up its members, e.g. for the UK cabinet.

But now imagine a hypothetical world where Kim has seized power and assigned every cabinet role to themself. Then, as you put it, there is only one actual individual that represents the cabinet – Kim is the whole cabinet!

In many cases, we can say that plurals have parts, which are singular. But in the general case, knowing what is a “part” is difficult, and we can’t tell from the syntax alone.

Got it. Thanks @guyemerson and @Dan. Just like many of the the other predications I’ve written the logic for, it depends on the context and the model. OK, I’ll play around and see if I can get something working.

As a related question: I have been interpreting _part_q_of in the above MRSs as a “generic” quantifier that doesn’t do anything beyond get the set of things defined by its RSTR and apply them to the BODY with no “extra” semantics. I.e. it doesn’t have “extra” semantics like requiring a certain number of things (like “the” does), or pick an arbitrary one (like “a/an” does), etc. Is that right? or are there extra semantics that part_q_of is meant to imply?

Edit: I ask because I’ve noticed that sometimes part_q_of is not generated in “…all of the X” sentences. For example: “look for all of the fuses” doesn’t have it. That’s what made me start to wonder if there is some extra semantic it is trying to capture:

[ TOP: h0
INDEX: e2
RELS: < 
[ _the_q LBL: h14 ARG0: x13 [ x PERS: 3 NUM: pl IND: + ] RSTR: h15 BODY: h16 ]
[ _fuse_n_1 LBL: h17 ARG0: x13 [ x PERS: 3 NUM: pl IND: + ] ]
[ pronoun_q LBL: h4 ARG0: x3 [ x PERS: 2 PT: zero ] RSTR: h5 BODY: h6 ]
[ pron LBL: h7 ARG0: x3 [ x PERS: 2 PT: zero ] ]
[ _all_q LBL: h9 ARG0: x8 [ x PERS: 3 NUM: pl ] RSTR: h10 BODY: h11 ]
[ part_of LBL: h12 ARG0: x8 [ x PERS: 3 NUM: pl ] ARG1: x13 ]
[ _look_v_for LBL: h1 ARG0: e2 [ e SF: comm TENSE: pres MOOD: indicative PROG: - PERF: - ] ARG1: x3 ARG2: x8 ]
>
HCONS: < h0 qeq h1 h5 qeq h7 h10 qeq h12 h15 qeq h17 > ]

There is part_of and _part_q, but no _part_q_of or part_q_of.

part_of is an abstract predicate (also called a grammatical predicate or gpred) – it is a “part” as a technical linguistic term (i.e. a word in the metalanguage).

_part_q is a real predicate corresponding to the English word “part” (i.e. a word in the object language). It is also a quantifier.

If you compare “we ate {none, some, part, most, all} of the rice”, all the MRSs contain part_of, but only the one with “part” contains _part_q.

(As a side note, “generic” also has a technical meaning, referring to “generic sentences” like “ducks lay eggs”. So “the generic quantifier” actually refers to a specific quantifier, namely the one used in generic sentences.)

[Sorry, I was being sloppy with my mental pattern matching there by suggesting there was a _part_q_of as opposed to just _part_q]

Got it, thanks for clarifying, @guyemerson.