Neuro-Fuzzy Approaches to Natural Language Data Processing
Ulrich H. Langanke · 2007
Dubbed the 'age of permissiveness' by traditional mathematicians, philosophers and linguists, the period since the late 1960ies has marked a fundamental change in the estimation of logical exactness of the type 'A or not-A'. Since then, the realm of Aristotle, the cultural exclusiveness of binary logic, has been seriously challenged by Fuzzy Logic. Fuzzy sets, vagueness, multivalence and multivalued categories of the type 'A and not-A', have caused paradigm shifts in those technical sciences focussing on the simulation of the human neural network, e. g. cognitive sciences, artificial intelligence, informatics, mechatronics, automation and language processing. However, they do not seem to be equivalently considered in product oriented linguistics, despite some theoretical efforts in cognitive linguistics. Regardless of typological differences such as print-based or electronic, even recently released dictionaries, maps/atlases exploring data on language diversity, language learning programmes or language processing are caught in the 'memetic' trap of bivalence. This essay reveals the interdisciplinary methodological interdependence between the three levels of 1) programming and language processing, 2) the linguistic data set as a continuum and 3) the design/presentation of fully fuzzified electronic language data sets based on the combined principle of Neuro-Fuzzy systems.