An Aggressive Robin In The Backyard: Formal Quantification Of Prototypicality Level Within The Frame Of The Prototype Semantic Theory Of Cognitive Linguistics

It is argued that an interdisciplinary approach to description of linguistic phenomena combining formal and cognitive linguistic theories provides a foundation for a research and scientific platform greatly facilitating exchange of valuable specific knowledge, benefitting different involved scientific domains, in this case computer science and cognitive linguistics.This paper proposes a means of introducing formal methods for assessing the level of representativeness within the frame of the prototype theory of cognitive semantics. The methodology and terminology are taken from fuzzy logic theory and linear algebra.A model is proposed wherein categories are represented as fuzzy sets containing concepts. Their membership is not of binary character as proposed by radical structural semantics; instead, it is quantitified by concepts' properties which contribute in various amounts to their representativeness within the pertaining categories. For each property of a concept a function needs to be defined by which the contribution of the property to the membership to the fuzzy set is quantified. The proposed model posits concept representativeness as a vector within a vector space such that each space dimension corresponds to one property. Coefficients along basis vectors that span this vector space comprise the weight factor of the pertaining property and the value of the function quantifying the membership of the concept to the category with respect to that property.Furthermore, the paper proposes the application of the described model to complex semantic structures. The representativeness of a situation within the pertaining situation category is a vector in of a vector space space spanned by basis vectors which in this case correspond to concepts constituting the complex semantic structure. Situation representativeness is a function of selection restrictions posited to each constituting concept by other concepts within the same complex semantic structure and representativeness of each concept within its category if observed isolatedly.After offering the description of the quantification model, it is argued that such a model may provide a basis for the improvement of existing or devising of new computer systems for natural language processing by implementing the principles of categorisation undisputedly present within the cognitive processes. Finally, areas for further research and improvements are suggested.