Mystery Number Two - Computational Linguistics and Talking Robots - page 33

3. Mystery Number Two:
Natural Language Communication Cycle

The second mystery is the compositional semantics of natural language. How
are lexically analyzed word forms connected into content in the hear mode?
How is content activated selectively in the think mode? How is activated con-
tent realized as external surfaces in the speak mode? To answer these ques-
tions, let us begin by further refining the format of analyzed word forms.

3.1 Choosing the Data Structure

From a software engineering point of view, choosing the format and the inter-
pretation of the basic units for modeling the cycle of natural language com-
munication amounts to choosing a certain data structure. 1 A well-defined data
structure is required for specifying the input and output interfaces of the algo-
rithm intended to operate on the data items of the system, as in their storage,
processing, and memory retrieval.
The items recognized and produced by the cognitive agent's language inter-
faces are the external surfaces. Therefore, internal word forms are the natural
candidates for being defined as the basic data structure. The format, called
proplet , develops from more intuitive representations:

3.1.1 D EVELOPMENT OF THE PROPLET FORMAT

2.1.1 2.6.1 lexical proplet

surface level: eau sur: eau surface

meaning

internal
surface

noun: water

convention

cat: sn f morpho−syntactic

properties

sem: mass
fnc:
mdr:
pc:
nc:
prn:

meaning level:

lexical

syntactic−semantic
connections
to other proplets

analysis

analyzed

word form

1 In early computer science, the term “data structure” referred to hardware constructs such as queue,
stack, linked list, car, cdr, array, or pointer. The development of higher programming languages