Mystery Number Five - Computational Linguistics and Talking Robots - page 93

The final proplet of a motion sequence is characterized formally by its empty
next value, just as the initial proplet has an empty prev value (6.1.2, 6.1.3,
6.1.4, 6.1.7). The Word Bank proplets activated by the navigation of the LA-
act1 grammar are used as blueprints for the agent's action components.
LA-act1 is simple because the concatenation between proplets is limited to
the continuation attributes prev and next . Also, building the hardware for the
core values red, green, blue, strght, left, and right should not be too difficult.
The LA-act1 system illustrates the basic DBS constructs listed in 1.1.2.
LA-act1 resembles the propositional calculus of Symbolic Logic in that
both model coordination . 3 They differ, however, in that propositional calcu-
lus is designed to define the truth conditions of formulas like (p r , while

LA-act1 is designed to model the behavior of an agent in the form of stimulus-
response sequences. Accordingly, the constants and variables of standard bi-
valent propositional calculus have only two semantic values, namely true (T,
1) and false (F, 0), while the semantic values of LA-act1 proplets comprise an
open number of recognition and action procedures, such as those defined in
6.1.5.

∧ q) ∨

6.2 Guided Patterns to Expand a Fixed Behavior Repertoire

Conceptually, the interaction between a Word Bank and an LA-act grammar
may be compared to the interaction between a (vinyl) record and a record
player. 4 However, while the signals on a record are stored in a time-linear
sequence in the record groove, the temporal dimension is recoded in a Word
Bank by means of certain proplet-internal values, allowing the signals to be
stored in alphabetically ordered token lines. The temporal order of a sequence
is realized by the rule applications of the LA-act grammar, which drive the
focus along the semantic relations between proplets, coded in the continuation
attribute (here next ) and the book-keeping attribute 5 (here prn ).
Providing a fixed behavior agent with additional stimulus-response pairs re-
quires translating them into sets of proplets (as in 6.1.2-6.1.4), which are
stored in the agent's Word Bank. Thus, it is not an extension of the LA-act rule

3 See Hausser 2003 for a reconstruction of propositional calculus in DBS, with special attention to
Boolean Satisfiability (SAT).
4 This two-level structure is one of several differences between the formalism of LA-grammar and
formalisms based on possible substitution, such as Phrase Structure Grammar, including Finite State
Automata. For a comparative complexity analysis of LA-grammar and PS-grammar, see FoCL'99,
Chaps. 8-12. For an analysis of the relation between LA-grammar and FSAs, see CoL'89, Sect. 8.2.
5 For a more detailed explanation of the different attribute kinds of a proplet, see NLC'06, Sect. 4.1.