direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Analysis of Error Responses in Human-Speech-Assistant-Interaction

Location:  TEL, Room Auditorium 3 (20th floor), Ernst-Reuter-Platz 7, 10587 Berlin  

Date/Time: 09.12.2019, 14:15-15:00 

SPEAKER: Sihan Yuan (TU Berlin)

Abstract: Errors heavily affect user experience in interaction with Voice User Interfaces (VUI). However, there is limited research on how error responses in VUI affect User Experience (UX). Therefore, I studied user preferences of different error response categories, and evaluated user experience for different error responses. My study consisted of three stages. In the first stage, 61 participants took an online survey to categorize different error responses. In the second stage, 91 subjects participated in a sorting survey and ranked different error response categories according to their own preference. In the third stage, the most popular error responses were used in a Wizard of Oz (WoZ) experiment to analyze how responding time affects user experience during voice interaction. The results of my studies indicate that neutral acknowledgment of an error is preferred by participants over emotional or apologetic responses. Furthermore, my findings suggest that there is no significant difference in user experience between the group in which timing of audio responses was based on emotion detection and the other two groups in which timing was fixed (after 0 or 2 seconds). It suggests that emotion detection might not improve UX when used to time error responses. Combined these results indicate that emotion detection might not improve user experience or liking when used for content generation or for timing of error responses. My work provides some initial findings regarding content and responding time of error messages of speech assistants, which may contribute to the field of error response design in Voice User Interfaces (VUI).

 

 

Zusatzinformationen / Extras

Direktzugang

Schnellnavigation zur Seite über Nummerneingabe