Building a Controlled Natural Language Framework for Real-time Machine Translation

Abstract

This study presents a framework that addresses the need for real-time communication between persons without a common language through the use of Controlled Natural Languages (CNLs). The framework consists of a set of CNLs (initially based on English and Spanish) where text compliant with any one CNL is mapped to an internal, unambiguous, language-in- dependent representation which, in turn, is automatically translated into valid text in a target natural language that not only retains the meaning of the original text but also appears natural to native speakers. In addition, the study proposes a software system similar to modern instant messaging clients to effectively facilitate said communication.