Usability of Automatic Speech Recognition Systems for Individuals with Speech Disorders: Past, Present, Future, and A Proposed Model

Abstract

People are using voice assistants (VAs) such as Siri & Alexa more than ever before. With 46% of U.S. adults using VAs, commercially available voice-activated technologies are becoming pervasive in our homes and beyond (Pew Research, 2017). VAs provide convenience, novelty, and unique solutions for the medical industry. But, some users may be left out of the conversation. People with speech disorders or atypical speech historically have found difficulty with using automatic speech recognition (ASR) technologies, the precursor to VAs. Usability testing for these systems has consistently shown that they are not easy to use for people with speech disorders. This investigation sought to perform a literature review of the existing research on the usability of commercially available ASRs for people with speech disorders to provide historical perspectives and to take an inventory of how this issue is being addressed today. A literature review was performed on the usability of commercially available ASRs for people with speech disorders and was divided into two stages: studies before the introduction of VAs and those that tested VAs themselves. Understanding where we have been and where we are now will also inform technical communication and usability professionals on what the future of ASRs may hold and how we can best address the needs of this audience. To do so, this paper proposes solutions for inclusive design in the voice assistant design space including a conceptual model for integrating specific techniques into commercially available VAs.