A semiotically oriented cognitive model of knowledge representation

This thesis introduces a model for knowledge representation as a sign recognition process, on the basis of an analysis of the properties of cognitive activity. By offering a logical account of this model, the existence of a `naive' logic underlying human information processing is revealed, which in turn opens the way towards a Peircean semiotic characterization of the cognitive model. `Naive' logic is a procedure generating relations between collections of qualia, in the sense of agreement, possibility, and (relative) difference. It is suggested that those relations have common meaning aspects shared with Boolean relations on two variables. The close relationship between the process model of cognitive activity on the one hand, and the Peircean signs on the other enables the cognitive model to be interpreted as a meaningful process, and the Peircean classification of signs as a process, generating meaning aspects or parameters of (meaningful) interpretation. In conformity with the fundamental nature of cognitive activity, it is suggested that the process model of cognitive activity may be uniformly applied for modeling different knowledge domains. This hypothesis is tested for the domain of `naive' logical, syntactic, semantic syntactic, reasoning and mathematical symbols. Each of these models consists in a specification of a recognition process (parser) and a definition of combinatory properties of primary entities (lexicon). An advantage of the proposed theory is that adjustments of the model of a domain, for example, in order to cope with new phenomena, may only require an adjustment of the lexicon, not the parser, which can be invariantly used. An advantage of uniform knowledge representation is that it may reduce the hard problem of merging complex signs obtained in different domains to the more simple task of structural coordination. Such a representation is used in this thesis for the definition of a technique for text summarization.