Voice-based Conversational User Interfaces (CUIs or VUIs) are being rapidly adopted by service providers and other organisations seeking to automate interactions with customers and clients. Indeed, it is becoming increasingly rare to speak with a human agent when contac0ng an organisation. Ensuring that these ‘conversations’ are successful is challenging both for the conversation designers and engineers developing the technologies underpinning the CUI; and to anyone else required to engage with the CUI. To enable this, engineers rely on a range of technologies, such as speech recognition and Natural Language Processing and Understanding (NLP/NLU), to ensure the system understands the speech of the human client, and to genera te speech that is understandable and appropriate for the type of activity. This becomes exponentially more challenging in interactions where the client’s speech deviates from that on which the machines have been trained, including that of L2 speakers. Such clients may face situa0ons where their speech is unrecognised or wrongly processed by the machine. In the absence of a human interlocutor who may be able to adjust to this, this could lead to false outcomes, disengagement and marginalisation. This presentation reports on work carried out by the authors in partnership with healthcare start-up Ufonia. The company has developed a voice assistant, Dora, for carrying out routine clinical conversations across a number of hospital trusts (see Brandt et al. 2023). Using in- sights and methods from Conversation Analysis, the collaboration has worked to ensure that these conversation-framed user activities achieve the institutional aims of the phone-call, while providing an acceptable experience for the user. Specifically, here we report on how the recent adoption of Large Language Model (LLM) technology for carrying out these activities has opened up further opportunities to leverage Applied Linguistics to help train the CUIs in recipient design for L2 users. More information about the project is available here: https://interactionalai.com/
Speaker bio: Spencer Hazel is an Associate Professor in Applied Linguistics & Communication at Newcastle University. Hazel’s research deals with how people interact, both with each other, and with technology such as conversational user interfaces (CUI). Hazel adopts an conversation analytic approach to do so. Currently, he is pursuing research in the use of conversational AI, international workplaces, dementia communication, and theatre settings.
Hazel previously worked as a researcher at the University of Southern Denmark, as research associate for the projects CALPIU (Cultural and Linguistics Practices in the Internationalised University, Roskilde U), LINGCORP (Language and Interaction in the Global Corporate Workplace), and Transient Multilingual Communities, collaborating with researchers from Roskilde University, University of Copenhagen and Copenhagen Business School. Hazel has a background in the performing arts and in language teaching.