28.10.2014 Views

Empirical Issues in Syntax and Semantics 9 (EISS 9 ... - CSSP - CNRS

Empirical Issues in Syntax and Semantics 9 (EISS 9 ... - CSSP - CNRS

Empirical Issues in Syntax and Semantics 9 (EISS 9 ... - CSSP - CNRS

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Class Subj<br />

export: e<br />

identities: e= 0<br />

syntactic dimension<br />

S<br />

NP [I= 1][AGR= 2] VP [AGR= 2]<br />

[E= 0]<br />

NP ≺ VP V⋄<br />

semantic [ dimension ]<br />

event<br />

0<br />

EFFECTOR 1<br />

Class DirObj<br />

export: e<br />

identities: e= 0<br />

syntactic dimension<br />

VP<br />

V⋄ [E= 0] NP<br />

[I= 1]<br />

semantic [ dimension ] [ ]<br />

event ∨ event<br />

0<br />

0<br />

GOAL 1 THEME 1<br />

Class IndirOjb<br />

export: e<br />

identities: e= 0<br />

syntactic dimension<br />

VP<br />

V⋄ [E= 0] NP<br />

[I= 1]<br />

V≺NP<br />

semantic ⎡ dimension ⎤<br />

causation<br />

[ ]<br />

0⎢<br />

⎣<br />

change-of-poss ⎥<br />

⎦<br />

EFFECT<br />

RECIPIENT 1<br />

Figure 10: MG classes for subject, direct object, <strong>and</strong> <strong>in</strong>direct object<br />

other argument realizations are possible as well <strong>and</strong> should be taken <strong>in</strong>to account <strong>in</strong> the metagrammar<br />

classes. For <strong>in</strong>stance, the subject NP class Subj should not only conta<strong>in</strong> the base subject<br />

realization shown on the left of Fig. 10 but also a tree fragment for an extraposed subject, for a<br />

wh-extracted subject, for a relativized subject, etc. Some of these tree fragments will contribute<br />

different aspects to the semantics. We leave this aside for the moment, s<strong>in</strong>ce the focus of this<br />

paper is on the dative alternation <strong>and</strong> its semantics. In this paper, we treat only the active base<br />

case, assum<strong>in</strong>g that other cases can be captured along the l<strong>in</strong>es sketched <strong>in</strong> Fig. 5.<br />

Let us first consider the classes needed for the DO construction. Some of the classes are just<br />

small tree fragments that do not use any other class. These are, for <strong>in</strong>stance, the ones for the<br />

different arguments, namely, for the subject NP, the direct object NP, <strong>and</strong> the <strong>in</strong>direct object NP.<br />

The first two are fairly general; they occur <strong>in</strong> many of the elementary trees <strong>and</strong> do not constra<strong>in</strong><br />

the semantics. The three argument classes are shown <strong>in</strong> Fig. 10. Only the roles relevant for<br />

our constructions are given, there are of course more possible roles for these arguments. Each<br />

class has a name, a declaration of variables that one can refer to when us<strong>in</strong>g this class (the export<br />

variables), a list of equations, <strong>and</strong> a syntactic dimension <strong>and</strong> a semantic dimension. The syntactic<br />

dimension conta<strong>in</strong>s a tree description that is depicted <strong>in</strong> the usual way <strong>in</strong> the figure. That is, solid<br />

l<strong>in</strong>es <strong>in</strong>dicate immediate dom<strong>in</strong>ance, dotted l<strong>in</strong>es <strong>in</strong>dicate dom<strong>in</strong>ance <strong>and</strong> the order of sisters<br />

<strong>in</strong>dicates l<strong>in</strong>ear precedence (but not necessarily immediate l<strong>in</strong>ear precedence). Furthermore, ≺<br />

denotes immediate l<strong>in</strong>ear precedence. In the class Subj, for <strong>in</strong>stance, the tree description tells us<br />

that there are three nodes n 1 , n 2 , n 3 with labels S, NP, <strong>and</strong> VP such that n 2 has a top feature I with<br />

value 1 . Furthermore, n 1 immediately dom<strong>in</strong>ates n 2 <strong>and</strong> n 3 (depicted by the edges of the tree)<br />

<strong>and</strong> n 2 immediately precedes n 3 . The representation is a bit sloppy s<strong>in</strong>ce it mixes node variables<br />

with node categories. The realization of the third argument as an NP (i.e. the use of the class<br />

IndirObj) is responsible for the caused possession mean<strong>in</strong>g. Therefore this class contributes a<br />

frame fragment <strong>in</strong> its semantics that tells us that the mean<strong>in</strong>g is a causation whose effect is a<br />

change of possession where the argument contributed by this class denotes the recipient. 9<br />

Concern<strong>in</strong>g the semantic dimension, we assume this to be a description of a typed feature<br />

structure. When we say ‘unification’, speak<strong>in</strong>g of comb<strong>in</strong><strong>in</strong>g frames <strong>in</strong> the metagrammar, we<br />

9 Aga<strong>in</strong>, this is of course not the only way this syntactic fragment can be used; other semantic contributions of<br />

<strong>in</strong>direct objects must be specified <strong>in</strong> the metagrammar as well.<br />

179

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!