Mystery Number Five - Computational Linguistics and Talking Robots - page 95

These proplets are lexical because only the core attributes have values, here
the attributes rec (for recognition) and act (for action) .
Once the fixed behavior agent has been extended to recognize the action
steps of certain guided responses, the lexical lookup triggered by guiding the
agent through the steps of 6.2.1 will result in the following sequence:

6.2.4 R ECOGNITION AND LEXICAL LOOKUP OF MOTION PATTERN 6.2.1
⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

rec: red
prev:
next:
prn: q 1

rec: green
prev:
next:
prn: q 2

act: strght
prev:
next:
prn: q 3

act: left
prev:
next:
prn: q 4

act: left
prev:
next:
prn: q 5

act: left
prev:
next:
prn: q 6

act: left
prev:
next:
prn: q 7

To complete this sequence into the proplets of 6.2.2, the core values must be
cross-copied to the prev and next slots of adjacent proplets. This is done by
the following LA-rec1 grammar, which uses the same variable definition as
LA-act1 (6.1.5):

6.2.5 R ULE SYSTEM OF THE LA- REC 1 GRAMMAR

{( rec: T n {Rule_0, Rule_1}) }

ST S = def

Rule_0 {Rule_0, Rule_1}

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦ ⇒

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

rec: T n
next:
prn: K i

rec: T n +1
prev:
prn: K i +1

rec: T n
next: T n +1
prn: K i

rec: T n +1
prev: T n
prn: K i +1

Rule_1 {Rule_2}

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦ ⇒

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

rec: T n
next:
prn: K i

act: M1
prev:
prn: K i +1

rec: T n
next: M1
prn: K i

act: M1
prev: T n
prn: K i +1 v

Rule_2 {Rule_2}

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦ ⇒

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

act: M1
next:
prn: K i

act: M2
prev:
prn: K i +1

act: M1
next: M2
prn: K i

act: M2
prev: M1
prn: K i +1

{( next: rp Rule _ 2 )} ST F = def

During a derivation, the parser ensures that all instances of K match the same
letter value.
The LA-rec1 grammar 6.2.5 is completely general. Given the variable defini-
tion 6.1.5, it handles all possible guided patterns using the stimuli red, green ,
and blue and the response steps strght, left, and right . If there is no restriction
on the length of the different sequences, their number is infinite. 8 Furthermore,

8 In a concrete agent, the number of stimulus-response pairs actually implemented will always be finite,
thus requiring only a fixed amount of memory.