Joshua Crowgey

Part I

Dan:

JC:

Part II

Questions

Antske:

JC:

Antske:

Guy:

JC:

EMB:

Antske:

Guy:

David:

JC:

David:

JC:

David:

JC:

David:

JC:

David:

JC:

Woodley:

JC:

Dan:

JC:

Dan:

Francis:

JC:

Guy:

JC:

Back up scribe's notes

Antske: Your questions about causatives reminds me about the event extraction community --- yearly workshop at NAACL. Going with annotation guidelines that ask people to put things on a timeline. Two separate things v. one annotated that way. Something to look into, but not sure if we want to take that over. TimeML always does separate events. For RED guidelines they don't do that. Not sure yet if the guidelines are openly published yet. Making the causative the same event is appealing, but you can't have entities playing a different role having the same role label for the same event.

Joshua: But different PRED values! Isn't that enough, to give different interpretations of ARG1?

Antske: If I were to go to a completely semantic representation, I'd say there's one event with these roles, and then it becomes unclear who's doing what.

Guy: Why not add an ARG2 with the causativizing suffix, keeping the same predicate symbol. Just unify in an ARG2. Leaving the other underpsecified.

Emily: Linking to ARG1 and ARG2 in that case? Do we want to interpret ARG1 as undergoes and just have the added argument (agent) be ARG2?

Antske: Do all verbs do that, or just a group of verbs? Could say they're ARG2 only, until the ARG1 gets unified in.

Guy: Ann's already said she doesn't want to attach any interpretations to specific ARG1.

David/Antske: Can the unergative ones go through causative?

Emily: Is it the same causative?

Joshua: Not clear.

Guy: Does the internal causative only attach to things that are very patientive to begin with?

Joshua: I've done some empirical grepping, but no generalizations stood out. Each verb is attested with a couple different ones of these; moribund language, can't do elicitation.

David: On the second issue of NPs. Do you have to have a determiner for all NPs?

Joshua: I wonder that too. Mostly yes, aside from proper nouns.

David: Do names get determiners?

Joshua: Not usually.

David: Because this determiner can be put in front of anything, what happens when you put it in front of something that's been causativized, and does it differ depending on the determiner? If you treat the NPs as basically verbal, I think there are some problems with that. There are still reasons to think that NPs are underlyingly nouny, but it's extremely subtle.

Joshua: You can tell the ones that are nouny v the ones that are verby by their morphotactic potential.

David: Possessive marking only on nouns?

Joshua: Yes.

David: Even a determiner in front of a VP can't have a possessor?

Joshua: Right -- that's one of the main criteria for determining a noun from a verb.

Woodley: Does the causative morpheme go on non-verbs?

Joshua: Transitivizers only attach to verbs; nouns can take subject but can never become transitive.

Woodley: Prepositions, adjectives?

Joshua: Only two prepositions: one locative and one grammatical.

Emily: I'm wondering if we can Dan to expand on his reaction…

Dan: Getting use to the idea of the same variable showing up as ARG0 in two EPs, but surprised by the repeat of the ARG1. Why not splurge and use ARG2?

Joshua: Predicate-symbol specific interpretation.

Dan: What work it the event doing for you, other than serving as a bookkeeping device?

Joshua: What more work should I want it to do?

Dan: Tense/aspect -- stuff that can happen to one of the events but not the other if they are distinct? Negation negating one but not the other? It seems like it ought to be possible to determine empirically whether the grammar treats them as separate?

Joshua: That only goes one way, right? If I don't find evidence that they're separate, then what do I know?

Dan: I was hoping you'd find some.

Dan: As soon as you use ARG2, why not just use one EP?

Joshua: I've got multiple different causatives -- internal, external, out of control, in control. Encoding that with the predicate symbol.

Emily: But those could be intersective modifiers…

Francis: What would make the whole discussion stronger is some way of telling the different. The intuition that this describes it better is hard to justify. The one you've seen first or in a nice presentation will make more sense. Are there any linguistic tests/things where sharing an ARG0 makes a different prediction?

Joshua: The only types of things I know about so far are things that, if you find them, would argue for more than one ARG0. Having a single ARG0 constitutes a hypothesis that you won't find them.

Guy: But you've still said two predicates, not just one.

Joshua: I'm still not sure what it means for something to be a predicate.

StanfordCrowgeyIntrinsicVariableNotes (last edited 2016-06-16 23:11:11 by EmilyBender)

(The DELPH-IN infrastructure is hosted at the University of Oslo)