WeldVUI: Establishing Speech-Based Interfaces in Industrial Applications

Voice User Interfaces (VUIs) and speech-based applications have recently gained increasing popularity. During the past years, they have been included in a wide range of mass-market devices (smart phones or technology installed in common car cockpits) and are thus available for many everyday interaction scenarios (e.g., making phone calls or switching the lights on and off). This popularity also led to a number of guidelines for VUI design, software libraries and devices for speech recognition becoming available for interface designers and developers. Although generally helpful, these resources are often broad and do not fully satisfy the specific requirements of certain industrial applications. First, grammar and vocabulary in such settings usually differ drastically from everyday scenarios. Second, common software libraries and devices are often not able to comply with the conditions in industrial environments (e.g. involving high levels of noise). This paper describes the iterative, user-centered design process for VUIs and functional speech-based interaction prototypes for the domain of industrial welding, including a two-stage Wizard of Oz procedure, rapid prototyping, speech recognition improvement and thorough user involvement. Our experiences throughout this process generalize to other industrial applications and so-called "niche applications" where grammar and vocabulary usually have to be established from scratch. They are intended to guide other researchers setting up a similar process for designing and prototyping domain-specific VUIs.