Practical Challenges of Virtual Assistants and Voice Interfaces in Industrial Applications

Abstract

Virtual assistant systems promise ubiquitous and simple access to information, applications and physical appliances. Their foundation on intent-oriented queries and support of natural language makes them an ideal tool for human-centric application. The general approach to build such systems as well as the main building blocks are well-understood and offered as off-the-shelf components. While there are prominent examples in the service sector, other sectors such as the manufacturing and process industries have nothing comparable. We investigate the practical challenges to build a virtual assistant using a representative and simplified case from the domain of knowledge retrieval. A qualitative study reveals two major obstacles: Firstly, a high level of expectations from users and, secondly, a disproportional amount of effort to get all details and having a robust system. Overall, implementing a virtual assistant for an industrial application is technical feasible, yet requires significant effort and understanding of the target audience.