Discourse analysis of the artificial intelligence assistant

Abstract

The primary purpose of this study is to analyze the components and classify the functions of conversational discourse of the A.I. assistant. Basically, language is a medium for interaction, so to analyze the discourse of conversation can be to analyze of what human and A.I. do with language in a social context. In this study, discourse is viewed as a language use that constitutes and frames a way of talking. Generally, the addresser (mostly human) and the addressee (mostly artificial intelligence assistant) exchange messages in a common context. In this situation, the contact is offered by a face-to-screen method. According to the modified communication model by structural linguist Roman Jakobson’s, types of discourse can be classified by language functions. According to the analysis, the artificial intelligence assistant basically performs the referential and conative functions at this point, and specific functions are prominent, depending on the service type. For example, Siri is an ‘empathy-oriented’ type that performs the emotive and phatic functions, in addition to the referential and conative functions. Google Assistant is an ‘information transfer’ type with enhanced referential and conative functions, while Cortana is a ‘relation-oriented’ type, in which the phatic function overlaps. The speech environment is analyzed through visualization elements strengthened for each service through interface design based on Gestalt theory. It is impossible to perform the metalinguistic function at this stage, but the development of discourse between human and A.I. Assistant will be expanded in various ways through advances in technology. The result of this study can be expected to contribute to a basis for future communication skill in the post-human era. © 2018 SERSC Australia.