Database Reference
In-Depth Information
theories capable of accommodating continuous upscaling at a nontrivial level
are verified to be suitable for further development.
Therefore, the long-term success of upscaling the overall system must be a
major concern in today's linguistics. For empirical support, an efficient imple-
mentation is a methodological necessity because without it the relative cor-
rectness of the theory cannot be tested automatically and objectively on large
sets of relevant data.
9
After implementations in LISP and C, DBS is currently
implemented as the JSLIM software written in Java.
Of course, we can only hope that the design of DBS will sustain the up-
scaling effort long-term. We are optimistic, however, because the cognitive
distinctions and building blocks listed in 1.1.1 - 1.1.4 are so general, obvious,
and simple that an alternative model of natural language communication doing
without them is likely to have serious flaws.
1.3 Declarative Specification
For theoretical and practical reasons, we present the DBS theory of natural
language communication as a
declarative specification
in the sense of com-
puter science. In contradistinction to an
algebraic definition
in logic (which is
based solely on set theory
10
), a declarative specification may describe a soft-
ware machine in terms of such higher level notions as interfaces, components,
functional flow, data structure, memory, and algorithms (cf. AIJ'01).
A declarative specification abstracts away from the
accidental
properties of
an implementation (program), such as the choice of the hardware, the pro-
gramming language, and the idiosyncrasies of the programmer. Instead, a
declarative specification defines the
necessary
properties of the software and
of the theory on which it is based. As a consequence, one declarative speci-
fication may have an open number of different implementations which are all
equivalent with respect to the necessary properties.
11
9
This method of verification differs from those of the natural sciences (repeatability of experiments)
and mathematics (proof of consistency), though it is compatible with them; cf. NLC'06, Sect. 1.2.
10
Pure set theory is not equipped to represent the computationally relevant aspects of a software ma-
chine. High level modeling languages like UML (Universal Modeling Language, e.g., Fowler 2004)
and ER (Entity Relationship, e.g., Chen 1976, Beynon-Davies 2003) developed in computer science
are not flexible and detailed enough to address all the issues of DBS - though reconstructing certain
aspects of the theory in UML, ER, or other projects of similar intent could well be worthwhile.
11
The dichotomy between a declarative specification and its possible procedural implementations dif-
fers from the Nativist dichotomy between competence (language knowledge) and performance (pro-
cessing knowledge). The latter distinction is used for “competence grammars” intended to model an
innate human language ability, reified as a language acquisition device (LAD).
In this connection it is worth noting that Chomsky, Sag, and others have emphasized repeatedly
Search WWH ::
Custom Search