REX: General-Purpose CNL with Code Generation Support

Abstract

Controlled natural languages (CNLs) have been proposed to address some of the issues of natural language when it is used to express requirements. CNLs, however, are based on formal grammar, which can easily become complex, hard to read, and especially hard to write, and the implementation of support tools can also demand a significant effort. Moreover, unanticipated constructions cannot be handled or have to be handled in unexpected and cumbersome ways. In this article, we present REX, a CNL with a simple grammar that is, thus, easy to understand and easy to support, but still general purpose. To accomplish this, instead of trying to support every conceivable construction and imposing a language on the users, through a small but comprehensive set of rules and through patterns, users specify their own language and how natural it is. Another of the benefits of CNLs is the possibility to automate the transformation of a text or specification into something useful, thereby reducing manual labor and transformation errors. In this article, we also present the support tools that were used to transform a REX text into code and a complete application. It is also shown that this CNL and its support tools can be easily adapted to suit different needs.