Evaluation of a Speech Interactive CALL System

Evaluation of a Speech Interactive CALL System

Hazel Morton (University of Edinburgh, UK), Nancie Davidson (University of Edinburgh, UK) and Mervyn Jack (University of Edinburgh, UK)
DOI: 10.4018/978-1-59904-895-6.ch013
OnDemand PDF Download:
$37.50

Abstract

This chapter describes the design of a speech interactive CALL program and its evaluation with end users. The program, SPELL (Spoken Electronic Language Learning), combines speech recognition technology, embodied animated agents and virtual worlds in creating an environment in which learners can converse with virtual characters in the target language in real-time. Empirical results are presented from an experiment which evaluated user satisfaction of using the program and interacting with the characters, and speech recognition accuracy. The results detail the limitations of the speech recogniser in this context. However, data on user satisfaction clearly show that the system was perceived by language learners as being both useful and enjoyable.

Key Terms in this Chapter

CALL (Computer Assisted Language Learning): An approach in language learning in which a computer presents material to the learner or where the computer is used as a tool to aid language learning.

Recast: Recasts are a reactive, implicit form of corrective feedback, specific to the individual learner, whose original meaning is taken and restated by the interlocutor with some aspect (phonological, syntactic, lexical etc) being modified.

Animated Agents: Animated agents are graphical representations of characters used in computer applications, also referred to as Virtual Agents or Animated Characters. Animated agents which are used in conjunction with ASR are often referred to as Embodied Conversational Agents (ECAs).

ASR (Automatic Speech Recognition): A technology which allows users of an application to speak utterances in order to interact in or control some aspects of the application. Speaker dependent recognition systems require the user to train the system first with their voice; speaker independent systems are trained on large samples of user data and can be used without the individual user training the recogniser first.

Usability: According to the ISO standard, usability of a system is evaluated in terms of effectiveness (the extent to which the intended goals of use are achieved), efficiency (the resources, such as mental effort, that have to be expended in order to achieve the intended goals) and satisfaction (the extent to which the user finds the use of the product acceptable) with a particular system in a particular context.

Complete Chapter List

Search this Book:
Reset