Representing knowledge in controlled natural language: A case study

Abstract

In this case study I argue for the usage of a machine-oriented controlled natural language as interface language to knowledge systems. Instead of using formal languages that are difficult to learn and to remember for non-specialists, authors should be able to write specifications texts in a well-defined subset of English that can be unambiguously processed by a computer. This subset of computer-processable English (PENG) consists of a restricted grammar and lexicon and is used together with an intelligent text editor that guides the writing process. The editor of the PENG system communicates with a language processor that generates logical structures while the author writes a specification text. The language processor is connected via a server with reasoning services that allow for acceptability checking and question answering. Texts written in PENG look seemingly informal and are easy to write and understand for humans but have first-order equivalent properties. © Springer-Verlag 2004.