Database Reference
In-Depth Information
1. Introduction: How to Build a Talking Robot
Can computational linguistics steer clear of artificial cognitive agents with lan-
guage, i.e., talking robots? The answer is
no
if the research goal of our inter-
disciplinary field is a functional reconstruction of free natural language com-
munication, verified by a computational model. The practical outcome of this
approach is unrestricted human-machine communication in natural language.
A talking robot requires language cognition as well as nonlanguage cogni-
tion. For example, a human telling the artificial agent what to do requires the
machine to understand natural language and to perform nonlanguage action.
Similarly, for the artificial agent to tell a human what is going on requires the
machine to have nonlanguage recognition and natural language production.
1
The essential contributions of the robot metaphor to computational linguis-
tics are the need (i) for a distinction between the robot-external environment
and the robot-internal cognition, (ii) for interfaces providing a connection be-
tween the robot's environment and cognition, and (iii) for an autonomous con-
trol connecting the robot's recognition and action in a meaningful way.
2
We honor these plain facts by making them founding assumptions of Database
Semantics (DBS). They provide valuable requirements and restrictions which
differ from the founding assumptions of Symbolic Logic in analytic philos-
ophy and of Nativism in theoretical linguistics. They also provide excellent
heuristics for designing the functional flow of the DBS software.
1.1 Universals
DBS defines successful language communication succinctly as a transfer of
information from the cognition of the speaker to the cognition of the hearer,
solely by means of a time-linear sequence of modality-dependent unanalyzed
external language surfaces (Chap. 2). The transfer is successful if the informa-
tion coded by the speaker is reconstructed equivalently by the hearer.
1
Cf. FoCL'99, Sect. 23.5, for the ten S
LIM
states of cognition.
2
There are additional requirements, such as an agent-internal database, but these are not robot-specific.
