Computing Perspective in Text - Computational Linguistics and Talking Robots - page 223

This, however, is not even the main problem of Truth-Conditional Semantics
for building a robot. The fatal inadequacy is a methodological one: definitions
in a metalanguage cannot be understood by a computer. What is needed instead
are procedural definitions, suitable for implementation.
For example, Every farmer snores would be analyzed as
1.

snore(x)].
This expression is formally interpreted relative to a set-theoretically defined
model @ and a variable assignment g , superscripted at the end as in
2.

∀ x[farmer(x) →

snore(x)] @ ,g .
According to the metalanguage definition of the universal quantifier, formula
(2) is true with respect to the model @ if the formula without the universal
quantifier, 7 i.e.,
3. [farmer(x)

∀ x[farmer(x) →

snore(x)] @ ,g ,
is true with respect to the model @ and all assignments g to x - usually
infinitely many. This method rests entirely on understanding the meaning of
the word all in the metalanguage. 8
The DBS alternative is metalanguage-independent patterns which character-
ize all as exhaustive, some as selective, and so on (cf. NLC'06, 6.2.9). Using
the pattern names as proplet values, the “quantifiers” are redefined as deter-
miners. Instead of restructuring the content of 11.5.4 beyond recognition (and
without success), as in 11.5.5, the DBS alternative is an automatic, time-linear,
surface compositional hear mode derivation. The quantifiers binding variables
horizontally, causing the scope problem in 11.5.5, are set aside and replaced
by addresses which establish coreference between it and donkey successfully:

→

11.5.6 DBS GRAPH ANALYSIS OF THE D ONKEY SENTENCE

(i) semantic relations graph (SRG) (iii) numbered arcs graph (NAG)

beat beat

1 8

6

7

farmer (donkey)/ça farmer (donkey)/ça

2 5

own own

4

3

donkey donkey

7 The purpose is to reduce the interpretation to the truth conditions of propositional calculus.
8 Other reasons why Truth-Conditional Semantics is unsuitable for building robots are the
[-sense, -constructive] ontology on which it is based (cf. FoCL'99, Sect. 20.4), and the resulting ab-
sence of recognition and action procedures as well as a memory in an agent.