The symbol grounding problem and the genesis of semantic categories

• Grant number: 11/10286-3
• Support Opportunities: Scholarships abroad - Research
• Start date: September 15, 2011
• End date: August 14, 2012
• Field of knowledge: Linguistics, Literature and Arts - Linguistics - Linguistic Theory and Analysis
• Principal Investigator: Marcos Fernando Lopes
• Grantee: Marcos Fernando Lopes
• Host Investigator: Stevan Harnad
• Host Institution: Faculdade de Filosofia, Letras e Ciências Humanas (FFLCH). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: Université du Québec à Montréal (UQÀM), Canada

Abstract

Cognitive sciences and, particularly, semantics in cognitive linguistics, are concerned with two central research themes: categorization and the symbol grounding problem. Categorization isclosely linked to the notion of class, although it is unlikely, from the standpoint of human cognition, that lexical classes (for example)are formed exactly as classical mathematical sets, ie based on bivalent semantics. It seems empirically appealing to think of relative degrees of inclusion or exclusion in sets, as well as considering that some elements may fit their classes more than others, standing out as prototypes. The symbol grounding problem, on the other hand, relates to the fact that the meaning of symbols in any symbolic system cannot be based solely upon other (meaningless) symbols (Harnad 1990). This project focuses on human descriptions of the semantically restricted domain of objects in the field of view, therefore dealing with grounded symbols, in order to determine which attributes are typically and minimally needed for the establishment of a logical ontology of the described objects. In other words, what and how many features would be necessary and sufficient for individuation, categorization and comparison of objects. And what the relevant consequences are for the description when categories are generated by bivalent, trivalent or plurivalent semantics. Answers to these questions will form the theoretical basis of computational rules to be implemented in "Project Dalgarno", a multidisciplinary research project (currently in early stage of implementation) in collaboration with the University of São Paulo School of Engineering related to creating robotic systems capable of visual acquisition and description of objects in natural language. (AU)