Learners’ Feedback Regarding ASR-based Dictation Practice for Pronunciation Learning

Abstract

Although early ASR-based dictation programs were criticized for lack of accuracy and explicit feedback for L2 pronunciation practice, teachers and researchers have shown renewed interest. However, little is known about student reactions to ASRbased dictation practice. This qualitative study examines student perspectives, identifying advantages and challenges to working with dictation software and generating ideas for the ideal ASR dictation program. Advanced ESL participants (n=16) worked with Windows Speech Recognition in a three-week hybrid pronunciation workshop. The study identifies many themes, including advantages such as ease of use, usefulness for pronunciation learning due to feedback provided, and heightened awareness of pronunciation issues, but also disadvantages, such as frustrating levels of recognition, particularly in the first attempt, doubts of the program's transcription abilities, and lack of convenience. Participants reported that convenience and greater support in pronunciation practice would be important for an ideal program.

Effects of Automatic Speech Recognition Software on Pronunciation for Adults With Different Learning Styles

“…These dictation ASR lack built-in functions for language learning, and were developed mostly for native speakers' communication. However, they do not require a paid subscription and are therefore more accessible for learners, teachers, institutions, and researchers (S. McCrocklin, 2019b). This accessibility allows their developers to significantly increase speech recognition accuracy through a collection of vast speech databases.…”

Revisiting Popular Speech Recognition Software for ESL Speech

“…Recent research has shown ASR based dictation for L2 pronunciation practice provides a range of benefits. Providing a sense of human intelligibility (Mroz, 2018), the transcripts can be utilized to identify probable errors at the word and sound level (McCrocklin, 2019b) and detect error patterns (McCrocklin, 2019c; Wallace, 2016). Further, learners using dictation have improved not only in segmental accuracy (Liakin, Cardoso, & Liakina, 2014; McCrocklin, 2019a), but also in overall intelligibility (Mroz, 2020).…”

“…Despite renewed interest, little research has examined whether dictation programs have improved for nonnative speech. Some studies have noted that students still report frustration from lack of recognition (Liakin, Cardoso, & Liakina, 2017; McCrocklin, 2019c) and Ashwell and Elam (2017) suggested accuracy of modern programs may not have improved substantially. Although ASR development companies perform extensive testing to improve recognition, assessments by such companies are likely driven to increase accuracy of transcription for all users.…”

SalukiSpeech: Integrating a New ASR Tool Into Students’ English Pronunciation Practice

“…Dictation programs were not developed for language learning; instead, they are often offered as accessibility software or promoted as a way for native speakers to create texts more quickly and efficiently. Recent research has shown a range of benefits to practicing with dictation programs, including increased noticing of probable errors at the word and sound level (McCrocklin, 2019b) and increased attention to segments with recurring errors (McCrocklin, 2019c;Wallace, 2016). Students who practice with ASR-based dictation show improvement of segmental accuracy, both for vowels and consonants (Guskaroska, 2020;Liakin, Cardoso, & Liakina, 2014;McCrocklin, 2019a;Park, 2017), Further, the transcript can support greater learner autonomy by providing feedback (McCrocklin, 2016) and a sense of human intelligibility (Mroz, 2018), which can be motivating for students (Mroz, 2020).…”

“…However, researchers have raised concerns that the transcript may not be sufficient or usable as feedback (Strik et al, 2008). Although McCrocklin (2019c) showed that learners felt they were able to make sense of the transcript as feedback, learners in McCrocklin (2019b) recommended building in greater support for dictation practice activities. SalukiSpeech (available at https://www.salukispeech.com and launched by McCrocklin, Fettig, and Markus [2021]) is a new tool that has been created by the authors to meet this need by providing flexible dictation practice with greater learner support.…”