Conceptual fuzzy sets as a meaning representation and their inductive construction

Fuzzy sets provide a strong notation for representing real world concepts which are essentially vague. However they have problems caused by the restriction of numerical membership functions, restriction of logical expression, lack of context dependency, etc. These problems relate to the representation of the meaning of a concept. In this article, we propose Conceptual Fuzzy Sets (CFS), a new type of fuzzy sets which conform to Wittgenstein's ideas (Philosophical Investigations, Basil Blackwell, Oxford, 1953) on concept meaning. A CFS is realized as an associative memory, combining a long-term memory and a short-term memory thus reducing the complexity of knowledge representation. In addition to solving the above problems CFS provide simple formula for knowledge representation and the procedure to use this knowledge. We introduce an inductive method for constructing CFS based on neural network learning. the effectiveness of CFS and of the learning method is illustrated through their application to the recognition of facial expressions. © 1995 John Wiley & Sons, Inc.