Exploring Syntactic Alignment in Spoken Natural Language Human-Computer Interaction.

Abstract

Psycholinguistic research has consistently shown that a person’s syntactic and lexical choices affect interlocutor utterance construction, effectively priming specific lexical and syntactic constructs, a concept termed alignment. This alignment is a dynamic process by which we converge on shared mental representations primed within dialogue affecting the words and grammatical structures we use when conversing with a conversational partner (Pickering & Branigan, 1998). Such an effect can be caused by conscious consideration of ones audience (mediated alignment or audience design) or through unmediated (unconscious) priming. Recent interest in this area has looked at this effect within human-computer interactions (Branigan, Pickering, Pearson, & McLean, 2010), although this research has mostly focused on text communication and semantic effects. Perceived quality of the system impacts on the amount of lexical alignment between computer-human interlocutors (J. Pearson, Hu, Branigan, Pickering, & Nass, 2006), thought to be mediated by a desire to enhance communication with the interlocutor because of users’ preconceptions of system function, so that assumptions of limited capability lead to stronger alignment. Although lexical alignment with computers has been researched in text based constructions, syntactic alignment in spoken dialogue with computer systems has not been explored. We report an experiment (currently underway) that explores whether users similarly show variations in syntactic alignment based on beliefs about system function, and whether alignment is linked to satisfaction in interaction. Specifically, we examine whether interaction with a truly advanced computer with natural language capabilities may lead to a shift in the user model of computer function away from perceived limitations of the computer system towards more natural conversation-based levels, and also a shift in the user model of computer function away from perceived limitations of the computer system towards more natural conversation-based levels.