Mystery Number One - Computational Linguistics and Talking Robots - page 31

2.6.2 I NTERFACE E QUIVALENCE P RINCIPLE
For unrestricted human-machine communication, the artificial cogni-
tive agent with language (DBS robot) must be equipped with the same
interfaces to the external environment as the human prototype.
This principle follows from the modality-dependency of the external surfaces
and the associated interfaces. The external interfaces of humans are concretely
given and are therefore susceptible to an objective structural and functional
analysis. Using a combination of the natural sciences and engineering, such
an analysis helps to reconstruct the external interfaces of the artificial agent.
Another equivalence principle, also preformulated in NLC'06, Sect. 1.5, ap-
plies to the processing of the surfaces by the artificial agent:
2.6.3 I NPUT /O UTPUT E QUIVALENCE P RINCIPLE
In natural language interpretation, the artificial agent must analyze the
modality-dependent unanalyzed external surface by
(i) segmenting it in the same way into parts (i.e., word forms) and
(ii) ordering the parts in the same way (i.e., in a time-linear sequence)
as the human prototype; and accordingly for language production.
Failure to establish segmentation and order correctly may disrupt communica-
tion. Designing a talking robot without integrating 2.6.3 would violate a basic
rule of science and compromise the agent's functionality from the outset.
The Input/Output Equivalence Principle extends the backbone of surface-
based information transfer, 2.6.1, from single word forms to complex ex-
pressions: for producing the external surface of a complex expression, the
speaker's cognition must provide a sequence of analyzed internal word forms;
for interpreting a sequence of unanalyzed external word form surfaces, the
hearer's cognition must reconstruct a corresponding sequence of analyzed in-
ternal word forms, at least to the extent of lexical lookup.
An external time-linear sequence as the output of language production and as
the input of language interpretation suggests a time-linear derivation order for
the speak and the hear modes. For the cognitive connection between the two,
i.e., the think mode, DBS adopts the time-linear derivation order as well. 33
The time-linear design, motivated by simplicity and efficiency, raises the
question: does it work empirically? So far, the DBS analysis of different gram-
33 In this way, a constant switching between (i) the time-linear computing of possible continuations in
the speak and hear modes and (ii) a computing of possible substitutions in the think mode is avoided.