Skip to content

ErgSemantics_Apposition

EmilyBender edited this page Jun 4, 2015 · 10 revisions

ESD Test Suite Examples

Abrams, the dog, arrived.
Abrams, garden dog, arrived.

Linguistic Characterization

A full noun phrase is followed by another nominal phrase (NP or Nbar) whose referent is the same as the first NP. This appositive nominal can be set off by commas, dashes, or parentheses, none of which introduce a reflex in the semantics. The apposition construction introduces a two-place relation 'appos' whose first argument is the index of the first NP (the head), and whose second argument is the index of the second, modifying nominal. The LTOP of the whole NP is the label of the 'appos' relation.

Motivating Examples

  • Parentheticals may be ambiguous: Abrams (the dog) arrived or My dog (no joke) was on TV

  • Titles may look like apposition, but are not: contrast professor Kim Smith with the professor Kim Smith, where the first is a title construction (see Compounding), while the second is apposition.

ERS Fingerprints

When the modifying nominal is a full NP, the apposition construction simply introduces the two-place appos relation, whose first argument is the index of the head NP, and whose second argument is the index of the second, appositive NP:

appos[ARG1 x1, ARG2 x2]
[ARG0 x1]
[ARG0 x2]

When the modifying nominal is a singular count Nbar lacking a determiner, as in the second reference example, the apposition construction introduces an implicit quantifier 'udef_q' to bind the index of the Nbar:

appos[ARG1 x1, ARG2 x2]
[ARG0 x1]
udef_q[ARG0 x2]
[ARG0 x2]

Interactions

Open Questions

This analysis introducing a two-place relation is problematic, since its label must be bound to something, yet cannot be identified with the label of the restrictor in the head NP (analogous to ordinary compounding), since that label is no longer visible when the appositive attaches to the NP. So by the semantic algebra, the 'appos' relation's label is identified with the LTOP of whichever predicate takes the full NP as its argument, but this is neither intuitive nor entirely successful, since that identification per the algebra is not yet fully implemented in the grammar. The near-term aim is to get rid of the 'appos' relation, instead making use of the recently introduced mechanism for expressing ICONS (constraints between individuals), having the construction add an 'eq' ICONS between the index of the head NP and that of the appositive nominal.

Grammar Version

  • 1212

More Information