Evaluating Voice User Interfaces in Ubiquitous Applications

Evaluating Voice User Interfaces in Ubiquitous Applications

Valéria Farinazzo Martins Salvador (Universidade Presbiteriana Mackenzie, Brasil), João Soares de Oliveira Neto (Universidade Presbiteriana Mackenzie, Brasil) and Marcelo de Paiva Guimarães (Centro Universitário Adventista de São Paulo, Brasil)
DOI: 10.4018/978-1-61520-843-2.ch003
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

In the current trend of applications going more and more ubiquitous, it is necessary to determine some characteristics, requirements and properties that must be assured in order that the application provides quality service to its users. This chapter describes a study on the evaluation of Voice User Interface (VUI) in Ubiquitous Applications and discusses some of issues which may impact the evaluation process when using the voice as a natural way of interacting with computers. The authors present a set of guidelines and usability principles that should be considered when developing VUIs for Ubiquitous Applications. Finally, they present the results of a case study which was performed in order to test and exemplify the concepts presented here.
Chapter Preview
Top

Introduction

Besides recognizing the user voice, Voice User Interface (VUI) systems are able to understand what the user says and to supply responses to these inputs, usually in real time. The state-of-the-art in speech technology already allows the development of automatic systems designed to work in real conditions (San-Segundo et al, 2005). Companies such as Philips, AT&T, and IBM have invested on the development of speech systems for restricted domains. On the other hand, the proliferation of computing into the real world promises the ubiquitous availability of computing infrastructure; it demands new paradigms of interaction that are more natural and inspired by constant access to information and computational capabilities.

The current knowledge on VUI comes from small contributions of research projects. These contributions propose an assessment for sub-systems developed in these projects, and try to generalize and make recommendations for the evaluation of VUIs, such as PARADISE (Walker et al, 1997), EAGLES (Gibbon; Moore; Winski, 1997) and DISC (Dybkjaer; Bernsen, 2000). However, the general tendency of using VUI is changing from applications that are built to be used in controlled environments to ubiquitous applications. Hence there is an increasing need to determine the principles of evaluating VUIs in this complex scenario.

So far, the principles of VUI evaluation were focused on specific issues, such as adequacy of the noise degree in the inputs of the system and the spontaneity of the interaction by voice. But, in a more complex environment, it is necessary to address other issues intrinsically related to ubiquitous applications, such as: if the communication through voice is a better choice for situations where one desires freedom for eyes and hands; the relevance of noise product by the environment where the applications is running; and privacy concerns (Abowd; Mynatt, 2000), (Dawkins et al, 2009).

According to Deng and Huang (2004), and Niculescu et al (2008), systems based on voice commands must face some challenges to achieve massive acceptance by the several sectors of the society:

  • Reduce the gap between what technology currently offers in terms of interface usability and what users want really need from voice-based interaction system;

  • Building robust systems in all possible acoustic environments: the voice recognition systems work well in quiet environments, but when the user is in a noisy environment, generally he/she can not use the system effectively, due to the great decreasing of the recognition error rate. For many years that was the main issue studied by researchers of voice recognition, from both universities and companies;

  • Need of putting into action systems for natural language: until now, when users interact with voice-based systems, they are aware that their partner is a machine. This is a failure in the voice recognition process, because, commonly, the user prefers a more natural style and the casual conversation;

  • Deal with language accents, slangs and regionalisms.

The main goals of this chapter are:

Complete Chapter List

Search this Book:
Reset