Compositional Semantics - Computational Linguistics and Talking Robots - page 119

7. Compositional Semantics
A computational theory of cognition requires (i) completeness of function ,
(ii) completeness of data coverage , and (iii) performance in real time . (i) is
essential for the success of long-term upscaling, (ii) for survival in the agent's
ecological niche, and (iii) for ensuring (i) and (ii) in daily performance.
Part I approached functional completeness by modeling the mechanism of
natural language communication in the form of a DBS robot with a body, a
memory, and external interfaces for recognition and action. Part II turns to
completeness of data coverage. The data to be analyzed are kinds of content ,
used in nonlanguage and language cognition alike.
7.1 Forms of Graphical Representation
Programming the cognition of a DBS robot requires the explicit definition of
content proplets and pattern proplets. The proplet format is needed for such
functional purposes as storage in a Word Bank, activation (retrieval), schema
derivation, matching, inferencing, navigation, and language production. How-
ever, for a simple, schematic overview of many different kinds of content, an
equivalent graphical representation will be more user-friendly and also open
the road to the insights and methods of graph theory.
The graphical representation of content in DBS is desigend for a conceptual
analysis of the semantic relations of structure. 1 These are functor-argument
and coordination, at the elementary, phrasal, and clausal levels (3.5.6). The
purpose of the representation is (i) the semantic analysis of different content
constructions and (ii) the surface production in different natural languages.
As a brief introduction to the issue at hand, let us compare the graphical
representations 2 of the subject-verb relation in (i) Phrase Structure Grammar
(PSG) , (ii) Dependency Grammar (DG) , and (iii) Database Semantics (DBS),
shown for the sentence Julia slept :
1 The semantic relations of structure are distinct from the semantic relations of meaning, such as syn-
onymy, antonymy, hypernymy, meronymy ,and holonymy ,aswellas cause and effect .Cf.Sect.5.3.
2 First published in Hausser (2010).