Voice-Driven Modeling: Software Modeling Using Automated Speech Recognition

Abstract

Voice-driven programming allows engineers to alleviate physical discomfort, pain, and injury. It also has the potential to be faster than typing and assist those with disabilities. While there are a number of solutions to voice-driven programming, Model-Driven Engineering (MDE) has yet to exploit this non-conventional but high-potential approach to software development. Standard convention in MDE practice involves creating software models using a traditional mouse and keyboard combination, or whiteboard sketch hardware. In this position paper, we introduce our vision and ideas for a Voice-Driven Modeling (VDM) approach. Our vision involves a framework that includes 3 phases: Speech Processing, Natural Language Processing, and Context Specific Modeling. We describe these 3 phases in this paper, which others can apply in their attempts to realize VDM. We additionally include our research plans for developing a VDM solution targeted at Simulink models and our early proof of concept capable of implementing several example commands. We establish the pertinence of this work through a survey that finds negligible work on VDM and highlights the potential impact this can have on the field of MDE as a whole. Specifically, it is our position that it can have a positive impact on modelers in general, modelers with disabilities, and domain experts not familiar with modeling. It is our hope that this work helps fuel research in this area, allowing for a new way to develop software models.