A layered controlled natural language for knowledge representation

Abstract

PENG-L is a framework consisting of a set of controlled natural languages designed for knowledge representation that has the potential to bridge the gap between fragments of natural language and formal languages. PENG-L is bases on a canonical grammatical schema that can be adjusted to define various layers of controlled natural language with different expressive power. The weakest layer of PENG-L is equivalent to propositional logic and offers natural language constructions for describing propositional statements. The next layer is equivalent to a version of description logic and provides natural language constructions to make assertional, terminological, and conditional statements tha can be unambiguously translated into function-free logic programs. The most expressive layer is equivalent to first-order logic and consists of a well-defined subset of natural language that can be unambiguously translated into first-order logic via discourse representation theory. Writting specification texts in PENG-L is supported by a look-ahead text editor that indicates after each word form that the author enters how the text can be continued. The text editor communicates with a language processor that generates the logical representation for the text incrementally and interacts with third-party reasoning services for acceptability checking as well as for question answering.