An Analysis of Dialogue Repair in Voice Assistants
Matthew Galbraith
TL;DR
The paper investigates how voice assistants handle dialogue repair when users employ an other-initiated repair cue, specifically the word 'huh?', in English and Spanish. It analyzes Google Assistant and Siri through three tasks—repair production, repair comprehension, and user acceptability—and classifies assistant responses into ten repair strategies. Findings reveal that the assistants do not replicate human-like 'huh?' repair and tend to interpret unintelligibly literally or fail to fulfill the request, while users prefer repair strategies that seek or provide relevant information (Strategy 5). The study highlights gaps between human-human and human-machine interaction in interactional language and emphasizes the need for cross-linguistic data and more human-aligned repair mechanisms in SDS to enhance usability and naturalness.
Abstract
Spoken dialogue systems have transformed human-machine interaction by providing real-time responses to queries. However, misunderstandings between the user and system persist. This study explores the significance of interactional language in dialogue repair between virtual assistants and users by analyzing interactions with Google Assistant and Siri, focusing on their utilization and response to the other-initiated repair strategy "huh?" prevalent in human-human interaction. Findings reveal several assistant-generated strategies but an inability to replicate human-like repair strategies such as "huh?". English and Spanish user acceptability surveys show differences in users' repair strategy preferences and assistant usage, with both similarities and disparities among the two surveyed languages. These results shed light on inequalities between interactional language in human-human interaction and human-machine interaction, underscoring the need for further research on the impact of interactional language in human-machine interaction in English and beyond.