Towards Generating Model-Driven Speech Interfaces for Digital Twins

The recent decade saw enormous advances of the capabilities of speech synthesis and speech recognition. While specific benefits depend on the individual applications, speech interfaces typically increase accessibility, enable "hands-free" and "no-screen" interaction, and often increase interaction speed and allow for more flexible usage patterns and increased multitasking, leading to higher user satisfaction. This paper presents a method to transfer these powerful benefits to the digital twin (DT) domain by automatically generating speech interfaces for DT applications. Our approach is based on Model Driven Engineering principles, where we automatically deduce command patterns from structural model information as used in, e.g., DT platform specifications. The speech interface generation is highly configurable and extendable, thus it can be applied to different DT models. The concepts behind the generator are generic as well, thus they can be ported to other applications and platforms. We validate our approach by applying it to two DT demonstration cases and provide a detailed description of the sketch interface configuration workflow.