Burghard Rieger:
In: Becker, J.D. (ed.): Agents, Communication, and Cooperation, [Lecture Notes in Artificial Intelligence] Berlin/Heidelberg/New York (Springer)
Anything we know or believe about the world can (more or less precisely) be communicated verbally. We do so by using words, forming sentences and producing texts whose meanings are understood to stand for, represent, or deal with the topics and subjects, the domains and structures in the real world they are meant to refer to. Natural language texts (still) are the most flexible and as that a highly efficient form to represent knowledge for and convey learning to others. Traditional approaches to the study of language understanding in CL and AI employ rule based formats of linguistic knowledge and symbol representations of world knowledge structures to model language processing by machine. Providing these initial knowledge bases and allowing them to be modified by system designers (external change), or dynamically as a function of processing (internal learning) proved to be everything from enormously laborious to error prone, from extremely difficult to virtually impossible. Computational Semiotics (CS) neither depends on rule-based or symbolic formats for (linguistic) knowledge representations, nor does it subscribe to the notion of (world) knowledge as some static structures that may be abstracted from and represented independently of the way they are processed. Consequently, knowledge structures and the processes operating on them are to be modelled procedurally and have to be implemented as algorithms which determine SCIP systems. As a collection of cognitive information processing devices these systems' semiotic character consists in their multi-level representational performance of (working) structures emerging from and being modified by such processing. The emergence of semantic structure as a self-organizing process ist studied on the basis of word usage regularities in natural language discourse, whose linearly agglomerative (or syntagmatic ) and whose selectively interchangeable (or paradigmatic ) constraints are exploited by text analysing algorithms. They accept natural language discourse as input and end up to produce a vector space structure as output. This may be interpreted as an (internal) representation of the semiotic system's states of adaptation to the (external) structures of its environment as mediated by the natural language discourse processed. In order to evaluate the internal picture which the system computes from the natural language texts according to its processing capabilities against the external reality whose structure and properties are described by natural language discourse only, a corpus of texts - composed of correct and true sentences with well-defined referential meanings - was generated according to a (very simple) phrase structure grammar and a fuzzy referential semantics which interpret simple composite predicates of cores (like: on the left, on the right | in front, behind ) and hedges (like: extremely, very, rather | nearby, faraway ). Processed during the system's training phase, the corpus reveals structural constraints which the system's hidden structures or internal meaning representations apparently reflect. Compared with a two-dimensional representation of the external reality - as described by the texts and specified by the underlying syntax and semantics - a two-dimensional transform of the system's internal view of its environment proves to be surprisingly adequate.
The system's architecture is a two-level consecutive mapping of distributed representations of systems of (fuzzy) linguistic entities whose states acquire symbolic functions that can be equaled to (basal) referencial predicates. Test results from an experimental setting with varying fuzzy interpretations of hedges will be produced to illustrate the SCIP system's miniature (cognitive) language understanding and meaning acquisition capacity without any initial explicit syntactic and semantic knowledge.
HTML Format
PDF Format (417 Kb)