Mystery Number Four - Computational Linguistics and Talking Robots - page 74

DBS inferences resemble LA-think rules like 3.4.2. However, instead of
merely navigating from one proplet to the next, an inference matches its an-
tecedent to a content in the Word Bank in order to derive new content by means
of its consequent. For example, using the format of pattern proplets and con-
tent proplets, the formal definition of the chain-initial R inference of 5.2.1 and
its application to a content is as follows:

5.2.3 F ORMAL DEFINITION AND APPLICATION OF A DBS INFERENCE
antecedent consequent

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦ cm

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

noun: β

fnc: hungry
prn: K

verb: hungry
arg:

noun: ( β K) verb: eat

arg: (

noun: food
fnc: eat
prn: K+1

rule
level

β
prn: K

fnc: eat
prn: K+1

K) food
prn: K+1

β

⇑ ⇓

matching and binding

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

noun: moi
fnc: hungry
prn: 211

verb: hungry
arg: moi
prn: 211

noun: moi
fnc: eat
prn: 211+1

verb: eat
arg: moi food
prn: 211+1

noun: food
fnc: eat
prn: 211+1

Wo rd
Bank

input output
The inference is activated by the content I am hungry (input), which matches
the antecedent. Utilizing the values to which the variables and K are bound

(i.e., moi and 211 , respectively), the new content I eat food isderivedbythe
consequent (output).
The inclusion of the antecedent's subject in the consequent by means of the
address value (

β

K) excludes cases in which one agent is hungry and another
one eats food - which would fail as an effective countermeasure. Because
repeated reference by means of an indexical is exempt from the coreference-
by-address method, the value corresponding to (

β

K) in the content derived by
the consequent is simply moi rather than (moi 211) . 8
The consequent of a DBS inference may contain variables which do not
appear already in the antecedent. An example is

β

in the consequent of step 4
in 5.2.1. This is permissable as long as the co-domains of such variables are
restricted to a certain range of possible values, as illustrated in 5.3.5.
The application of DBS inferences is governed by the following principle:

γ

5.2.4 S EQUENTIAL I NFERENCING P RINCIPLE (SIP)

Any two inferences x and y may be applied in sequence if, and only if,
the consequent of x equals the antecedent of y.

This basic principle provides for a kind of self-organization (Kohonen 1988).
Via equality of the antecedent or the consequent with other inferences, any