Human understanding of Controlled Natural Language in simulated tactical environments

Abstract

Computational platforms with natural language interfaces have become commonplace, but they present limitations that make them less than ideal for military and other safety-critical environments. Controlled Natural Languages (languages built from a subset of natural language, which are both computer- and human-readable) hold promise for Human-Computer Collaboration via these platforms, especially when the human user needs to add information to a knowledgebase or make queries, as they provide a transparent, shared representation. Controlled Natural Languages, however, are typically not optimized for human use and understanding. This paper presents the development and implementation of a framework to test the relative ease of comprehension of different Controlled Natural Language statements. The experiments presented in this paper show an advantage for one particular Controlled Natural Language statement over another, but only when responses are made under strict time pressure. These types of experiments allow researchers to make recommendations on how to improve the use and design of a Controlled Natural Language for more robust comprehension, particularly in tactical environments.