SBIR-STTR Award

This proposed project provides integrated methodology for developing natural language interfaces (NLIS). Using current lingustics (X-bar theory and augmented phrase structure grammars) and artificial intelligence (structures, frames and connections) research, it encompasses innovative (1) codeless-programming developmental tools for building NLIS, (2) the option of parallel processing for efficient semantic interpretation, and (3) a declarative programming approach to accurate disambiguation. 1. Linguistic computation diagrams (LCDS) are representations of data and control flow which naturally maintain data-typing and functional programming discipline. Special diagram macros called structure-gates (analogous to digital logic circuits) represent conditional actions connecting syntactic and semantic structure computations. 2. Speed of semantic interpretation would be increased by parallel computation. Until there is a parallel nli plug-in board, the diagram interpreter will be a serial simulation of the intrinsically parallel computation represented by an LCD. 3. Accuracy of real-time semantic interpretation could be the most important benefit of the prolog based declarative disambiguation technique, which attempts to eliminate hypothetical interpretations by searching in the knowledge structure for contradictions.