[FIXME --- put in summary of initial motivation]

oe: What does it mean for an eventuality to be [ SF ques ]?

EMB: When we got rid of the messages, we pushed the info down onto the event variables, but that doesn't mean it's really a property of them. MRS needs reinflating in that respect.

Guy: Proposes using ICONS to associate wh parameters with the appropriate [ SF ques ] bearing e.

Woodley: Aren't these about truth conditions as in John knows who saw what? And isn't ICONS only about non-truth conditional stuff?

Dan/Emily: ICONS will eventually also have constraints on anaphoric binding.

Woodley: Okay that's also truth conditions.

oe/Dan: What's true about ICONS so far is that it's all non-scopal, which is also true of this. Modeling only in terms of e & x.

Guy: Seems like the ERG doesn't allow multiple echo question examples.

Dan: That's something that I don't know how to do. Don't handle multiple gap sentences.

Woodley: Which sky would fall if you extended the max length of SLASH? Massive overgeneration or efficiency?

Dan: Both -- going all the way open. But changing the honest lie from there's at most one thing on the SLASH list to there's at most two would let us play with these examples and that might be reason enough to try it.

Guy: If we had appends for proper lists rather than diff-lists we would know the end (and thus when to stop popping), but that becomes messy.

oe: What's the summary of what needs to be captured in the semantics?

Guy/Emily: Equivalent of params list, ICONS seems to work.

Dan: Two-way distinction between wh words: Those that introduce a param lists element (on ICONS) and those that don't.

Emily: But the echo question ones also go on the params list, they just belong to the list associated with the top-most EP (e associated to ARG0 of EP with LTOP).

oe: Did Kim say who saw who? -- I thought that the params list would be empty, because we're not asking anything other than yes or no. But now I see that one flat params list doesn't do it. So we need something like binary relations ... like what's on ICONS.

Emily: Trying to get away from echo questions, working on constructing data from Mandarin:

Kim 说    谁     看   了  谁    吗
Kim shuo1 shei2 kan4 le  shei2 ma
Kim say   who   see  PRF who   Q

With ma -- only means 'Did Kim say who saw who?'

Without ma -- ambiguous: 'Who did Kim say _ saw who?' or 'Who did Kim say who saw _?' or 'Kim said who saw who'

Guy: So it's a strength of the current analysis for Mandarin that it's all underspecified?

Emily: No, because we need to be able to model the fact that with the ma the wh words must stay low. If we don't indicate somehow that the wh params only attach low, we've lost information that the grammar should give us information.

Emily/oe summary: When we got rid of messages, we lost the place to store the parameters as associated to a particular message/clause/question. So instead, now proposing to do this as binary relations on the ICONS list, with the e from the main verb as the stand-in for the clause.

oe: So what about scopal modification then? If we pick the event of the main verb, we've made a commitment.

Guy: G&S say that aren't scopally active ... but still somewhat? Who read every book? is absoultely ambiguous.

Guy/oe/Glenn: The things we need to worry about are scopal modifiers: negation, probably.

Guy: Who didn't leave isn't ambiguous --- so maybe it's fine to take the event variable of the main verb. G&S might have a story about Who read every book? that is other than ordinary scope ambiguity.

Woodley: How does attaching to a particular event variable say anything about scope?

oe/Woodley: ICONS aren't part of the scope tree.

oe/Guy: But using the event is saying we don't need the scope tree --- we're saying they are independent of the scope tree. In Who probably didn't leave? we don't need to say anything about scope.

Dan: If there were grammatical constraints, we'd be losing information.

oe: So we need to stack them, which is hard.

Guy: But even with just one, if there were ambiguity in Who never left? ... but I can't see what it is.

oe: Earlier you stacked questions to motivate the need for associating the params list with the clause. Now I'm asking whether the event variable is the right way to get that.

Emily: Worries some about the notion of 'clause' which is ill-defined for us and was part of the problem with messages (slippery slope).

Woodley: How black was the cat?

Emily: There you've used the construction that says that the wh word is unambiguously associated to the Matrix clause. To really test that, we'd need to embed the question in the nominal modifier, which I don't think English lets us do, except maybe in echo question land.

Dan: This slope isn't slippery: The wh words 'launch' the param-icons, and the grammar can have constraints on which constructions allow them to attach (meaning which event variables are available to them).

OsloQuestionRepresentation (last edited 2017-08-10 08:39:11 by EmilyBender)

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