New type (property) of variable?

Add property (y) in addition to event and ref-index? Not urgent. Can head-off criticism by saying at the beginning of the talk that eventuality will be pronounced event, and they are eventuality in a very broad sense comprising states of affairs and properties, in the sense of (Asher, some impenetrable book, page 538).

(Actually 1993)

p = underspecification of h and x, but only at the description level.

Would only show up in unrealized arguments. Which end up as i, because x would lead to non-scoping, and p would include handles.

Allow for unbound/unexpressed x?

Concern for unbound x is scoping machinery, currently in LKB, Utool

current hierarchy:

p := u.
i := u & 
 [ EXP exp/unexp ].
h := p.
x := p & i.
e := i.

Don't want to complicate the hierarchy, so use a property instead. The distinction is orthogonal to these. But we're already using multiple inheritance.

Currently used in the SEM-I only to mark arguments that can be either h or x (but not e). But it's not otherwise used in parsing or generation. But when the subject is dropped in Abandoning Kim arrived. the ARG1 of abandon is i, not p. A mystery. Must involve (stupendously undocumented) stupendous cleverness in the SEM-I.

Aim for fewer quantifiers?

No, except in as much as other changes to semantic analyses (e.g., putting fewer entities in MWE names) do so as a side-effect

WeSearch/HankøTheRest (last edited 2013-04-22 18:06:33 by EmilyBender)

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