Foundational Work in Glue Semantics
An LFG-based deductive approach to the syntax-semantics interface
This page focuses on work up to 2001.
Foundational Work in Glue Semantics
Formalisms for assembling a meaning for a natural language utterance
have traditionally been closely tied to syntactic structure. For
example, the Montague approach of function application presumes a
deeply nested binary tree structure to guide the
applications. Categorial grammar approaches combine both syntax and
semantics in the same formalism, leading to formidable complications.
We believe that the functional structure of Lexical-Functional
Grammar is better suited to provide the syntactic information
necessary for constraining derivations of meaning in a
cross-linguistically uniform format. It has been difficult, however,
to reconcile this approach with the combination of meanings by
function composition.
![[proof of `Bill left']](graphic.gif)
In contrast to compositional approaches, we have been working on a
deductive approach to assembling meanings, based on reasoning with
constraints, which meshes well with the unordered nature of
information in the functional structure. We found that simple logical
implication was very good at modeling the assembly of meanings in a
sentence, and could connect up with the functional structure
easily.
Schematically, if we have the sentence "Bill left", we can take
the semantic contributions to depend on syntactic information in the
following way:
-
- bill: The subject means Bill
-
- left: If the subject means X, then the sentence means left(X).
from which we can infer that the sentence means left(Bill).
Much more complex constructions can be handled straightforwardly. Our
use of linear logic as a `glue' for assembling meanings also allows
for a coherent treatment of modification, and the linear logic version
of implication is just what is needed to capture the property of
natural language that each part of a sentence contributes exactly once
(usually) to the final meaning.
Natural Language Theory and Technology
Palo Alto Research Center
3333 Coyote Hill Rd.
Palo Alto, CA 94304 USA
This page updated March 2006.