A semantic processor for natural language

Abstract

The problem of the extraction, representation, and utilization of meaning from English text has been elusive. Semantic parsing permits effective and relevant response to complex questions where the domain is well defined. The semantic net used is a three-row numeric matrix. Each column of the net represents an assertion about the domain. The meanings of each word are contained in associations (pointers to the names of things within the domain). The intersection of these pointers is used to define each phrase. Ambiguity is further reduced, when necessary, by an implementation of common sense-limiting the interpretations to those meanings that may be associated as the subject or object of each verb. Multiply embedded clauses are processed by passing the result of one clause as an argument for the next. By this means, complex expressions can be reduced to a single token.<>