Mystery Number One - Computational Linguistics and Talking Robots - page 32

matical constructions, such as relative clauses, gapping, bare infinitives, long-
distance dependencies, prepositional objects, copula constructions, etc. (cf.
Part II) in many different languages, has not discovered any empirical dif-
ficulty (fatal flaw) resulting from the use of a strictly time-linear derivation
order for the full cycle of natural language communication.
This is in contrast to Phrase Structure Grammar and Truth-Conditional Se-
mantics. Each has a fatal flaw when applied to natural language. Context-
free Phrase Structure Grammar cannot handle the syntactic phenomenon of
discontinuous elements, shown by Bar-Hillel (1953) (cf. FoCL'99, Chap. 8,
Constituent Structure paradox). Truth-Conditional Semantics is inherently
incapable of a complete analysis of natural language, shown by Tarski (1935)
based on the Epimenides paradox (cf. FoCL'99, Sect. 19.5).
In DBS, the Constituent Structure paradox does not arise because the lin-
guistic analysis is based on valency canceling in a time-linear derivation order
(cf. FoCL'99, 10.5.4), and not on constituent structure. The Epimenides para-
dox is avoided in DBS because meanings are based on the agent's recognition
and action procedures, and not on truth conditions.
Despite these long-standing results, Truth-Conditional Semantics and context-
free Phrase Structure Grammars continue to be the formal systems of choice
for today's mainstream approaches to natural language analysis. Simply ignor-
ing their flaws does not make them any less fatal, however. Instead, they go a
long way towards justifying the innovations and adaptations of DBS.
As the first, and so far the only, reconstruction of the cycle of natural lan-
guage communication, the agent-oriented approach of DBS is based on the
following innovations and adaptations:
1. the time-linear algorithm of LA-grammar,
2. the content-addressable memory of a Word Bank,
3. the data structure of proplets as non-recursive feature structures with ordered attributes,
4. the order-free coding of semantic relations between proplets by means of addresses,
5. the reconstruction of reference as an agent-internal correlation between language and context,
6. the reconstruction of intention in terms of maintaining the agent in a state of balance, and
7. the use of “ / ,” “ \ ,” “ | ,” and “ − ” edges to characterize functor-argument and coordination graphically.
These have created a divide between DBS on the one hand and the coalition of
sign-oriented schools of Nativism and their allies from Truth-Conditional Se-
mantics on the other. Yet the surface-based information transfer mechanism,
2.6.1, in combination with the principles 2.6.2 and 2.6.3 constitutes a simple,
effective, and natural method no freely communicating robot can do with-
out. Once accepted, the transfer mechanism sets into motion a chain of conse-
quences which are presented here as the software model of the DBS robot.