Inferences in natural language understanding

Abstract

This paper presents a subsymbolic fuzzy approach to language understanding. The author addresses the issue of natural language understanding in terms of two fundamental inferences: 1) the reasoning of implicit knowledge surrounding the text; and 2) the deduction from the explicit context of the sentences. Fuzzy reasoning and connectionist techniques are employed in order to yield right results. They are flexible enough to work in a natural language environment which is characterized by almost infinite variability. Firstly, linguistic meanings are encoded in subsymbolic microfeatures. A technique of fuzzy reasoning in a fuzzy rule matrix is used to capture the surrounding knowledge in the text. Finally, a fuzzy unification process is employed to form a coherent structure and comprehend the context.<>