Mystery Number Four - Computational Linguistics and Talking Robots - page 77

When this inference is applied to an input content like John had a good
meal , the result is the new content John had a meal which was not bad (cf.
6.5.9-6.5.11 for an analogous formal derivation). 9 Thus, the lexical meaning
relations defined in 5.3.1-5.3.5 are formulated once as inferences at the rule
level, but the results of their application may be written many times as new
content to the now front of the Word Bank.
The meaning relations are coded solely in terms of core values connected
by the usual semantic relations of structure, i.e., functor-argument and coordi-
nation. The new content coded by these relations may serve to trigger further
inferences, for example, for the interpretation of nonliteral uses. 10
Parallel to the meaning relations of lexicography are the event relations of
philosophy, most notably cause and effect . These events exist independently
of language, but due to the formal similarity between language content and
context content in DBS (4.3.3), event relations may be formalized in the same
way as the traditional meaning relations of lexicography:

5.3.3 I NFERENCE RULE IMPLEMENTING A CAUSE AND EFFECT RELATION
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤

noun: car
fnc: have
prn: K

verb: have
arg: car no fuel
prn: K

noun: no fuel
fnc: have
prn: K

noun: (car K)
fnc: no start
prn: K+M

verb: no start
arg: (car K)
prn: K+M

⎣ ⎦ ⎣ ⎦ ⎣ ⎦ impl ⎣ ⎦ ⎣ ⎦

This DBS inference formalizes the relation If a car has no fuel then it does
not start , where fuel may be instantiated by gasoline, electricity, methane, etc.
The descriptive power of the DBS inferencing method may be shown further
by condensing a complex content into a meaningful summary. As an example,
consider the following short text, formally derived in Chaps. 13 (hear mode)
and 14 (speak mode) of NLC'06:

The heavy old car hit a beautiful tree. The car had been speeding. A
farmer gave the driver a lift.

After syntactic-semantic parsing by LA-hear, the content of this modest text is
stored in the agent's Word Bank. A reasonable summary of the content would
be car accident . This summary may be represented in the agent's Word Bank
as follows:
9 The rules 5.3.1 and 5.3.2 illustrate how traditional meaning relations may be handled in DBS.
Whether or not the inverse relations hold as well is an empirical question which may be answered
by distinguishing between unidirectional and bidirectional inferences. Given that the inferences in
question are part of an individual agent's cognition, they do not have the status of universal truths.
Instead, they are evaluated and adapted in terms of their utility for the agent's day-to-day survival.
10 Thus, the hypernymy inference dog inst (antiates) animal may derive the new content Fido is an
animal from Fido is a dog . In the hear mode, this new content may be used to infer from the
language expression The animal is tired that the intended utterance meaning is Fido is tired (cf.
FoCL, Sect. 4.5; NLC'06, Sect. 5.4) - and conversely in the speak mode.