Open timjogorman opened 9 years ago
kevincrawfordknight
commented
9 years ago nice!
"One main thing about the AMR relationship to relative clauses is that we distin- guish between types of relative clauses (or mark them at all)..." Is there a "not" missing?
nschneid
commented
9 years ago Thanks for sharing! This sounds like a really cool event; I hadn't heard of it.
cbonial
commented
9 years ago Hi Tim,
Hope you're enjoying the meeting in Berlin. Some notes below on the annotations, overall wonderful job on these tough cases!
Claire
I was wondering why it appeared that some of the pronouns are inconsistent:
They forgot that they had voted.
(f / forget-01 :ARG0 (w / we) :ARG1 (v / vote-01 :ARG0 (t2 / they))) He relied on what they said.
(r / rely-01 :ARG0 (h / he) :ARG1 (t / thing :ARG1-of (s / say-01 :ARG0 (s2 / she)))) I personally like the "be-temporally-at" versions of this sentence: That they voted early surprised me.
(s / surprise-01 :ARG1 (i / i) :ARG2 (b / be-temporally-at-91 :ARG1 (v / vote-01 :ARG0 (t / they)) :ARG2 (e / early)))
I think the simpler version:
(all of these might be annotated as just:
(s / surprise-01
:ARG1 (i / i)
:ARG2 (v2 / vote-01
:ARG0 (t2 / they)
:time (e2 / early)))
)
may not capture the fact that it's the earliness, not the voting in itself, that was surprising.
But, it seems to me that "their voting early" is the arg0 cause, not the arg2 instrument of surprise.01.
Do you know from a larger context that this is an event, as opposed to a copula:
Who is?
(e / event-01 :ARG0 (a / amr-unknown))
Nice! Love this one: All Browne could do was arrive.
(i / include-91 :ARG1 (a2 / arrive-01 :ARG1 a) :ARG2 (t / thing :ARG2-of (c / capable-41 :ARG1 (a / animal :name (n / name :op1 "Browne")))) :ARG3 (a3 / all)) (representing: of the set of things Browne is capable of doing, Browne arriving fills that set) I'm a bit confused about "manage" and "manager". Is this a case where we decided to use the main predicate "manage" instead of have-org-role? It's inconsistently annotated in relation to have-org-role in the consensus instances of "manager":
Browne is a manager.
(h2 / have-org-role-91 :ARG0 (a3 / animal :name (n2 / name :op1 "Browne")) :ARG2 (m / manage-01)) Wouldn't "Brown arrived" be the ARG0 cause/reason, rather than the caused ARG1?
The reason: Browne arrived.
(c / cause-01 :ARG1 (a / arrive-01 :ARG1 (a2 / animal :name (n / name :op1 "Browne"))))
Do you know from a larger context that it is the students who are being evaluated? or are they doing the evaluations:
The student evaluations were negative.
(n / negative :ARG2-of (e / evaluate-01 :ARG1 (p / person :ARG0-of (s / study-01))))
I thought the standard for comparatives was to just put "cat" here, instead of repeating the property that's being compared:
The dog is older than the cat is.
(o / old :domain (d / dog) :degree (m / more :compared-to (o2 / old :domain (c / cat)))) For this one:
Abrams persuaded the dog to bark.
(p / persuade-01 :ARG0 (a2 / animal :name (n / name :op1 "Abrams")) :ARG1 (b / bark-01 :ARG0 (d / dog)))
the "persuade" arg1 is listed as the entity persuaded, while arg2 is actually the impelled action. Interesting though that this annotation could list the dog as arg1 and bark as arg2, or simply list the arg2 bark with the dog as the arg0 of barking.
I'm not sure I see a subevent relation here, they seem to be more sequentially related than an event within an event: Abrams left without paying.
(l / leave-01 :ARG0 (p / person :name (n / name :op1 "Abrams")) :subevent (p2 / pay-01 :polarity - :ARG0 p))
I feel like the purpose of the money should be captured here. To me, this seems like "the money used to buy the dog" arrived:
Money to buy the dog arrived.
(a2 / arrive-01 :ARG1 (m / money :ARG3-of (b / buy-01 :ARG1 (d / dog))))
Claire N. Bonial, PhD Student
Department of Linguistics Hellems 290 295 UCB University of Colorado Boulder, CO 80309-0295
On Nov 6, 2014, at 2:31 PM, timjogorman notifications@github.com wrote:
Hi all, Since I'm vaguely "representing" AMR at the comparative computational semantics meeting in Berlin, I figured I'd shared the webpage, which is now up with some our our analyses: http://moin.delph-in.net/WeSearch/Ccs .
The premise of the meeting is that we have a set of interesting examples we voted on ( http://svn.emmtee.net/trunk/uio/wesearch/ccs/berlin.txt), and will use that as a starting point for comparing all of these different representations. Some ones that seem interesting to AMR (my treatments are on the site) are things like:
Their voting early surprised me. Did you? All Browne could do was arrive. Nothing happened, which surprised him. She knew who hadn't arrived yet.
If anyone wants to look at these and has opinions, questions, or different analyses, feel free to post it here and I can try to relay that. Similarly, if anyone has any schemes that we might want to run by other people doing similar models (particularly scope/quantification), I can relay questions or proposals.
I'll report back in two weeks if we get any interesting conclusions.
— Reply to this email directly or view it on GitHub.
Hi all, Since I'm vaguely "representing" AMR at the comparative computational semantics meeting in Berlin, I figured I'd shared the webpage, which is now up with some our our analyses: http://moin.delph-in.net/WeSearch/Ccs .
The premise of the meeting is that we have a set of interesting examples we voted on ( http://svn.emmtee.net/trunk/uio/wesearch/ccs/berlin.txt), and will use that as a starting point for comparing all of these different representations. Some ones that seem interesting to AMR (my treatments are on the site) are things like:
Their voting early surprised me. Did you? All Browne could do was arrive. Nothing happened, which surprised him. She knew who hadn't arrived yet.
If anyone wants to look at these and has opinions, questions, or different analyses, feel free to post it here and I can try to relay that. Similarly, if anyone has any schemes that we might want to run by other people doing similar models (particularly scope/quantification), I can relay questions or proposals.
I'll report back in two weeks if we get any interesting conclusions.