Designing chatbots that produce language that is natural and appropriate to a given context is critical in satisfying user expectations. Currently, little is known about how a chatbot's linguistic choices should be designed to conform with the language humans produce in similar contexts. In this paper, we draw on existing sociolinguistic theory to adapt a technique calledregister analysis to (a) characterize the linguistic register used by humans in a specific conversational context; and (b) drive chatbot language design. Our exploratory study investigates the application of register analysis for tourist assistants chatbots and shows how the results could be used to develop them to adopt the appropriate register.
No takes yet. Share an insight, caveat, or question.
Chaves et al. (2019) studied this question.