A multimodal interface for controlled natural language processing

Over the last fifteen years, speech recognition has improved dramatically and sophisticated technology has become more available. It has encouraged a project that explores the integration of a speech interface in a controlled natural language (CNL) system, which prior to this, had yet to be explored. A CNL is a subset of a particular written natural language (base language) which has been engineered to restrict language components such as syntax, lexicon and/or semantics. By using a speech interface, API, this paper presents an implementation of a predictive editor that integrates text and speech as inputs in a CNL processing system, namely PENG-ASP Multimodal. In addition to building the system, a study was conducted to determine the practicality of a speech interface in a CNL system. In our findings, using speech input demonstrated significant benefits when looking at the speed during user input; however, the speed significantly declined as the number of errors during speech recognition grew.