CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No.

3; 2025

AI-Driven Pronunciation Assessment: The Impact of SpeechAce on EFL

Learners' Pronunciation Competency
Nguyen Ngoc Vu1, Vo Thiet Thach1, Tran Vu Diem Thuy1*
1
Ho Chi Minh City University of Foreign Languages - Information Technology, Vietnam
*
Corresponding author’s email: thuytvd@[Link]
*
[Link]
[Link]
®
Copyright (c) 2025 Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy

Received: 07/10/2024 Revision: 05/03/2025 Accepted: 11/03/2025 Online: 12/05/2025

ABSTRACT

The study examines the impact of SpeechAce, an AI-driven

pronunciation and fluency assessment tool, on enhancing students'
pronunciation skills. Utilizing a mixed-method experimental
design, the research involved an experimental group using
SpeechAce and a control group relying on traditional methods.
Quantitative data were collected through pretests, post-tests, and
questionnaires. Statistical analysis was conducted to compare the
effectiveness of SpeechAce with conventional teaching methods.
In-depth interviews were conducted to collect qualitative data on
participants' experiences and perceptions. The results revealed
significantly higher post-test pronunciation scores in the experimental
group compared to the control group. Additionally, the "students
perceived SpeechAce as a user-friendly and effective tool, particularly
praising its immediate feedback and interactive features.
Furthermore, SpeechAce's ability to pinpoint specific pronunciation errors
Keywords: improved students' performance. These findings have important
SpeechAce, implications for language educators and developers of educational
pronunciation technology, suggesting that integrating AI tools like SpeechAce can
competency, significantly enhance language acquisition and support the integration of
technology in ELT technology in English language teaching and learning.

Introduction
Pronunciation is an essential language component that learners need to acquire in order to
communicate effectively. Unfortunately, teaching pronunciation in Vietnam has been
overlooked since the ELT in the Vietnam education system favors receptive skills over
productive ones. The focus at the middle and high school levels tends to emphasize mainly on
grammar, reading, and listening, while productive skills, particularly speaking and
pronunciation receive less attention (Vu, 2016b). The curriculum is heavily centered on
exercises that develop reading comprehension and grammatical accuracy, with little attention
given to speaking or pronunciation. These two language components are also excluded from
the tests. Moreover, pronunciation is taught very briefly in the audiolingual method, which

CITATION | Nguyen, N. V., Vo, T. T., & Tran, V. D. T. (2025). AI-Driven Pronunciation Assessment: The
Impact of SpeechAce on EFL Learners' Pronunciation Competency. Computer-Assisted Language Learning
Electronic Journal (CALL-EJ), 26(3), 84-106. DOI: [Link]
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

emphasizes repetition and mimicry, offers only a superficial approach to pronunciation without
fostering a deep understanding or interest in the nuances of spoken English. Such a reptetitive
approach also demotivates students’ learning process. As a result, after seven years of learning
English, from grade 6 to grade 12, it is very hard for students to produce even the simplest
everyday conversation. This lack of focus on pronunciation has profound significant
implications: it not only limits students' ability to communicate effectively in English, but also
diminishes their overall motivation and enthusiasm for learning the language.
In contrast, an increased emphasis on pronunciation could potentially transform students'
attitudes towards learning English. By making pronunciation a central component of language
instruction, students could experience greater enjoyment and motivation in their language
learning journey. The rise of computer-assisted language learning (CALL) offers promising
new avenues for pronunciation instruction, particularly through the development of automatic
speech recognition (ASR) technologies. ASR has become an integral part of daily life,
embedded in virtual assistants like Siri, Google Assistant, and Cortana, which can understand
and respond to spoken commands. This technology has also found applications in education,
particularly in the realm of pronunciation training (Kumar & Renuka, 2023; Li, Deng, Haeb-
Umbach, & Gong, 2015).
One prominent example of ASR in pronunciation training is the software Praat, which analyses
human speech by generating detailed phonographic representations. These graphs visually
depict the pitch contours of speech, allowing learners to compare their pronunciation with that
of native speakers. By providing this visual feedback, Praat helps learners identify
discrepancies between their speech and the target pronunciation, facilitating more precise
practice (Hirai & Kovalyova, 2023; Jiang, Jong, Lau, Chai, & Wu, 2023). Similarly, other
learning English software which integrates ASR in their features such as Rosetta Stone,
Duolingo are also popular amongst English learners. Recently, there has been the presence of
SpeechAce, an ASR program that compares users’ speech to the native speaker’s ones. Such
ASR software like Praat and SpeechAce have paved new paths in seeking a better approach to
teaching and learning pronunciation. Thus, further studies need to be done on software similar
to those to introduce new methods in teaching pronunciation to students and make their learning
of pronunciation more motivating and meaningful.
In the context of global communication, English has emerged as a predominant language,
making proficiency in English a valuable skill. Pronunciation is a critical component of
language learning, as it directly impacts intelligibility and communication efficiency. However,
non-English-major students often face challenges in acquiring accurate pronunciation due to
limited exposure and practice opportunities. At a private university in Ho Chi Minh City,
catering to a diverse student body, many of whom are non-English majors, needs effective tools
and methodologies to enhance English language proficiency across its curriculum.
Nevertheless, non-English major students do not receive extensive pronunciation instruction in
the classroom. Pronunciation only takes a small part in each unit in the book English File (by
Oxford University Press). Moreover, formal pronunciation teaching is time-consuming because
instructors must address individual students’ varied pronunciation errors. As a result,
pronunciation is often taught reactively—only in response to noticeable mistakes—or

85
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

sometimes overlooked if the errors do not hinder intelligibility. This reactive approach has not
been effective, as students continue to repeat previously corrected pronunciation mistakes.
Given these challenges, ASR tools like SpeechAce could provide a valuable solution for this
university. By integrating such technology, students would have access to personalized and
consistent pronunciation practice, reducing the dependency on in-class instruction while
enhancing their overall language proficiency.

Literature review
Pronunciation in Language Learning
Pronunciation is a critical component of language learning, playing a vital role in a speaker's
ability to communicate effectively. According to Fraser (2000), clear pronunciation is essential
for ensuring that a speaker's intended message is understood by listeners without ambiguity or
misinterpretation. Correspondingly, pronunciation competency, as defined in the realm of
language learning, entails the ability to accurately produce the sounds of English, including its
vowels, consonants, and stress patterns, in a way that is intelligible to listeners (Celce-Murcia,
Brinton, & Goodwin, 2010). Derwing and Munro (2005) emphasize the importance of
intelligibility in pronunciation, highlighting that the primary goal is effective communication
rather than imitating native-speaker norms. Therefore, English pronunciation competency is a
multifaceted skill set crucial for successful communication in English, focusing on
intelligibility, comprehensibility and interpretability rather than adherence to native-like
pronunciation.
In the context of learning English as a second language, pronunciation greatly influences
intelligibility, which is the ease with which a listener can comprehend spoken words (Derwing
& Rossiter, 2002). Even when a learner's grammar and vocabulary are accurate, poor
pronunciation can lead to misunderstandings, thereby undermining communicative
competence, which is the ultimate objective of language acquisition (J. M. Levis, 2005).
Besides, acquiring accurate pronunciation poses several challenges for language learners,
particularly those whose native language differs significantly from English. One of the primary
obstacles is the phonetic disparity between the learner’s first language and English. For
instance, Vietnamese has a more limited range of vowel sounds compared to English, making
it difficult for Vietnamese learners to distinguish and produce English vowels correctly (Vu,
2024). This phonetic gap can create persistent pronunciation errors that are hard to overcome
without targeted instruction and practice.
Furthermore, many learners suffer from a lack of exposure to native English speakers and
authentic spoken English. This limited exposure hampers their ability to internalize the subtle
nuances of pronunciation and develop a natural-sounding accent (Gilakjani, 2012). The
situation is exacerbated by the fact that pronunciation is often neglected in educational
curricula, as is the case in Vietnam, where traditional language teaching methods prioritize
grammar and reading over speaking and pronunciation (Vu, 2016b). When pronunciation is
taught, it is frequently done using outdated methods that fail to engage students or address their
individual pronunciation challenges effectively (Derwing & Rossiter, 2002). Psychological

86
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

factors also play a significant role in the acquisition of pronunciation. Learners often experience
anxiety or embarrassment when speaking in a foreign language, particularly if they are unsure
of their pronunciation. This anxiety can lead to a lack of confidence and reluctance to practice
speaking, which further impedes their progress (Horwitz, 2001). Moreover, learners tend to
transfer the pronunciation rules of their native language to English, resulting in errors. For
example, Vietnamese speakers often struggle with the pronunciation of final consonant sounds
in English, which are less prominent in Vietnamese (Vu, 2016a).
Pronunciation proficiency has a profound impact on overall language learning outcomes.
Learners with good pronunciation are generally more confident in their speaking abilities,
which encourages more frequent practice and participation in conversations (Derwing &
Rossiter, 2002). This increased practice leads to greater fluency and competence in the
language. Additionally, accurate pronunciation enhances listening comprehension, as learners
become more attuned to the sounds and patterns of the language (Field, 2005). Conversely, poor
pronunciation can lead to frustration and discouragement, especially if the learner's spoken
language is frequently misunderstood. This frustration can decrease motivation and engagement
in language learning activities, ultimately slowing progress (Trofimovich & Isaacs, 2012).
Teaching and Learning English Pronunciation in Vietnam Context
In Vietnam’s public schools, English lessons mostly focus on grammar and skills like reading
and listening, while pronunciation tends to be overlooked. The current 7-year English program
doesn’t have specific sections for pronunciation practice—it usually just leaves students to learn
on their own using cassette tapes. But without proper guidance in the classroom, that kind of
self-study doesn’t really work. Because of packed schedules and pressure to cover the
curriculum, teachers don’t often have the chance to focus on pronunciation unless a student is
clearly struggling. On top of that, with big class sizes, old materials, and not enough time, it’s
tough for teachers to teach English effectively. Most students also rely heavily on the teacher’s
explanations and aren’t used to learning independently, which makes structured support for
pronunciation even more important.
Having realized the problem of the old curriculum, the Ministry of Education and Training
launched the National Foreign Languages 2020 Project, hoping to help students develop all four
skills equally and be able to use English to perform real-life tasks. Unfortunately, the project
was not a success. The Minister of Education and Training admitted that Project 2020 failed to
achieve its target. As acknowledged by the Minister of Education and Training, Project 2020
did not achieve its targets (T. Nguyen, 2017). Vietnamese students' English pronunciation
remains largely unintelligible to both native English speakers and even other Vietnamese
English speakers (H. N. Nguyen, 2023). This failure underscores the urgent need for greater
attention to pronunciation instruction and the adoption of new methods that can better motivate
students and enhance their communication skills. The lack of emphasis on pronunciation is
compounded by the fact that English teachers in Vietnam are generally not well-equipped to
teach this skill. A recent study by Loc and Newton (2020) revealed that most teachers approach
pronunciation instruction in unplanned and reactive ways, largely because their own training
did not include formal pronunciation classes. This was because they were not offered formal
pronunciation classes during their training. As a result, students’ errors were corrected with

87
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

certain repetitive techniques such as recasts or prompts. However, these types of feedback do
not have a long-term effect on students’ pronunciation competence. If overused, they might
demotivate the students’ pronunciation learning. They may also prevent students from
developing their autonomy. Thus, the students may never be able to learn pronunciation on their
own. The study also highlighted that teachers expressed a strong need for more formal
pronunciation training sessions to develop the skills necessary for effectively teaching
pronunciation to their students.
It is a common aspiration for learners of any skill to master all its components, and English
learners are no exception. As Ha and Bao (2023) noted in a study, students often aspire to
achieve native-like pronunciation or, at the very least, to be intelligible when speaking English.
The study also identified the common difficulties students face in learning pronunciation and
their desire for proper instruction that would help them express themselves more clearly and
become more autonomous in their learning. These findings align with those reported by Nguyen
(2017), emphasizing the importance that both teachers and students place on effective
pronunciation teaching and learning. Ultimately, intelligible pronunciation is key to motivating
students and enabling them to apply what they have learned effectively in communication.
However, integrating pronunciation instruction into an already-packed curriculum presents a
significant challenge. If there is insufficient time to include formal pronunciation lessons within
the existing schedule, alternative approaches must be explored to teach pronunciation
effectively without adding to the students' workload. This could involve the use of innovative
tools and methods that can be seamlessly integrated into the current curriculum, thereby
enhancing pronunciation instruction without overburdening teachers or students.
Automatic Speech Recognition for Pronunciation Teaching and Learning
As the name suggests, with automatic speech recognition (ASR), our speech is listened to and
analyzed by a machine for various purposes. For example, we can give commands to our phones
by talking to them using Siri for iPhones, Google Assistant for Android phones, or Cortana for
Windows laptops. The machines listen to our speech, analyze it, and execute the command
accordingly. ASR technology has also been applied in the field of teaching English
pronunciation through Computer-Assisted Pronunciation Training (CAPT) programs. Learners
record their utterances of words, phrases, or sentences into the CAPT program, which then
provides feedback focusing on the phonological features of their utterance. The types of
feedback and functions may differ depending on the programs. For instance, Spring and
Tabuchi (2022) discuss how ASR-based pronunciation practice can aid learners in improving
second language pronunciation, emphasizing the impact of treatment length and guided practice
on specific pronunciation points. Similarly, Moxon (2023) highlights the advantages of ASR
technologies, such as instant feedback and visual representation of sounds, which contribute to
improved intonation and pronunciation development.
One of the first and most notable program that used ASR in training pronunciation is Praat.
Developed by Paul Boersma and David Weenink of the University of Amsterdam, Praat
analyzes a learner’s speech in terms of intensity, pitch height, duration, or formants, then
compares the speech with a native speaker’s model through a display of speech waveforms.
Research on the application of Praat in pronunciation instruction has yielded promising results.

88
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

For example, Gorjian, Hayati, & Pourkhoni (2013) found that using Praat helped learners
acquire important prosodic features, including stress and intonation, which are critical for
effective pronunciation. More recently, a 2019 study by Spauling and Breazeal explored the use
of a pronunciation-based child robot, demonstrating that children’s literacy and pronunciation
skills improved after engaging in oral interactions with the robot. Similarly, ElsaSpeak, an
English-speaking practice app that utilizes ASR, has received positive feedback from learners,
who reported significant improvements in their pronunciation skills after using the app (Kholis,
2021; Sholekhah & Fakhrurriana, 2023). Despite the successes of ASR-based CAPT programs,
they are not without limitations. A key feature of these programs is their ability to provide
instantaneous feedback, which, while beneficial, also presents inherent drawbacks. There are
primarily two types of feedback in ASR pronunciation programs: visual feedback and ASR
feedback. Visual feedback, as provided by programs like Praat and Pronunciation Power,
displays the learner’s speech in a waveform graph, which requires specialized knowledge to
interpret. Without guidance from a teacher, learners may struggle to understand and apply this
feedback independently, limiting its effectiveness. On the other hand, ASR feedback is more
straightforward and basic. In programs like Rosetta Stone and Duolingo, learners’ utterances
are simply marked as correct or incorrect, with the degree of intelligibility often represented by
a visual indicator, such as a circle. However, these programs typically do not provide detailed
analyses of the phonological features of the learners’ speech, which could be crucial for
addressing specific pronunciation errors (Bajorek, 2017).
While CAPT programs that utilize ASR have shown promising results, significant challenges
remain. As J. Levis (2007) points out, "CAPT systems often suffer from difficulties in giving
learners adequate, accurate feedback and an inability to provide an accurate and automatic
diagnosis of pronunciation errors" (p. 185). In the following section, the discussion will focus
on an ASR-based CAPT program that has been used to teach pronunciation, with the hope of
addressing some of the gaps and limitations identified in previous programs.
SpeechAce: An Emerging Tool in Pronunciation Instruction
Given these challenges in existing ASR tools, SpeechAce has been developed to offer more
targeted feedback and greater accessibility for learners.
SpeechAce, developed by ex-Microsoft veterans Ian Cheung, Abhishek Gupta, and Ahmed El-
Shimi, is a speech recognition API designed for fluency and pronunciation assessment. Its
primary feature is the ability to score a learner's speech in real time, identifying mistakes at the
syllable and phoneme levels. The SpeechAce API supports a range of speaking activities,
including word and sentence-level pronunciation assessments, passage-level fluency
assessments, and the estimation of IELTS and PTE speaking scores (Moxon, 2021).
Additionally, it can transcribe and evaluate spontaneous responses, as well as assess grammar
and vocabulary. The SpeechAce API returns an overall score in each evaluated utterance and
detailed sub-scores providing feedback on mistakes at sentence, word, syllable, and phoneme
levels.
SpeechAce’s functions are straightforward and user-friendly for both the teachers who design
the tasks, and the users doing the tasks. First, teachers key in the prompt. After that, users turn

89
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

on the microphone on their electronic devices and record their answers. SpeechAce then
analyzes the speeches and gives feedback on users’ pronunciation. SpeechAce can be integrated
into various Learning Management Systems (LMS), such as Moodle, Blackboard, Canvas, thus
teachers can customize the tasks depending on the speaking skills they want learners to practice.
Teachers can also listen to learners’ speech and add their comments besides the comments
from SpeechAce. Users can practice their pronunciation endlessly, and they can also have
access to SpeechAce from different types of electronic devices. Additionally, the feedback is
comprehensive and consists of two main parts. The first part is the percentage score from 0 per
cent to 100 per cent on how close the user’s utterance is to the model one of the native speakers.
The percentage is also accompanied by comments such as “You got it! Are you a native
speaker?”, “You are not bad”, “Your pronunciation is off”. The second part is a detailed analysis
of individual syllables and phonemes of users’ utterances. Users can then compare their
pronunciation with the models. In addition to general feedback, SpeechAce also provides
IELTS scores for speaking, evaluating both fluency and pronunciation. Users can record their
responses to a prompt, receive an IELTS score, and review detailed feedback on their specific
mistakes, along with advice on improving their performance.
Figure 1
SpeechAce task and feedback

Compared to earlier ASR-based programs like Praat, Rosetta Stone, and Duolingo, SpeechAce
offers more specific and actionable feedback. While programs like Praat and Pronunciation
Power rely on complex waveform graphs that require specialized knowledge to interpret, and
others like Rosetta Stone offer only basic correct/incorrect feedback, SpeechAce provides
detailed phoneme-level feedback that pinpoints precise errors in pronunciation (Hassim,
Kassim, Kassim, & Kassim, 2023). This level of specificity is essential for effective language
learning, aligning with the pedagogical criteria suggested by experts like Pennington and Levis,
who emphasize the importance of setting measurable goals, providing consistent and immediate
feedback, and linking pronunciation to other aspects of communication.

90
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

Figure 2
Detailed analysis of pronunciation in SpeechAce

SpeechAce meets these criteria by offering pronunciation models in both American and British
accents, providing percentage-based scores that help learners set clear goals, and integrating
seamlessly with various LMS platforms. This flexibility allows teachers to design customized
tasks, and detailed feedback enables students to practice and improve their pronunciation at
their own pace, regardless of time or location. Although studies on the effectiveness of
SpeechAce are still limited, the existing research suggests significant potential for the tool to
enhance pronunciation skills. For instance, Moxon (2021) reported positive outcomes in
learners' pronunciation accuracy and self-confidence, though both studies noted limitations
such as the lack of control groups or balanced gender representation. Despite these limitations,
SpeechAce has demonstrated its value as a resource for addressing English pronunciation
challenges, particularly in educational settings.
Given the potential of SpeechAce and the gaps in existing research, this study aims to build on
previous findings by including both a control group and an experimental group, with an equal
number of male and female participants. By doing so, the study hopes to provide more robust
evidence of effectiveness in improving pronunciation and to encourage further research into the
application of ASR technology in pronunciation instruction. SpeechAce offers a promising
solution for pronunciation training, with the potential to make pronunciation teaching more
effective and accessible for both teachers and learners.
Research Gap
Despite growing interest in using Automatic Speech Recognition (ASR) tools like SpeechAce
for pronunciation training, existing research remains limited in several key areas. Many prior
studies have either lacked a control group or failed to ensure balanced participant demographics,
making it difficult to draw definitive conclusions about the effectiveness of such tools.
Additionally, while some studies have explored learners’ improvement in pronunciation, few

91
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

have systematically investigated learners’ perceptions of using ASR-based platforms in

instructional settings. This study seeks to address these gaps by incorporating both a control
and an experimental group, with an equal distribution of male and female participants, to offer
more reliable data. Furthermore, by examining not only pronunciation outcomes but also
learner attitudes, this research aims to provide a more comprehensive understanding of the
impact and user experience of SpeechAce in pronunciation instruction.
Research Questions
To fulfill the purpose of the study, the survey aimed to answer the following research questions:
1. To what extent does SpeechAce improve students’ pronunciation competency?
2. What are the students’ perceptions toward learning pronunciation with SpeechAce?
These questions seek to explore the effectiveness of SpeechAce in enhancing pronunciation
skills and to understand students' experiences and attitudes toward using this technology in their
language learning process.

Methods
Pedagogic Setting & Participants
The research site is a private university in Ho Chi Minh City, established in 1991. It offers a
diverse range of academic programs and serves a student body with varying levels of English
proficiency.. The study was done at the Department of General English, which belongs to the
Faculty of Foreign Languages. The department’s main program is called English for
International Communication (EIC). The program is designed to help students use English
confidently and effectively in daily contexts, in multi-cultural working environment, and in
academic settings. There are six levels in total, from EIC 1 to EIC 6. At the end of EIC 6,
students are required to take the Aptis test, which is developed by British Council. Students
have to obtain at least 120 points to successfully pass the level. The main course books
throughout the program are the English File 3rd Edition (Oxford University Press), from
elementary to intermediate plus. Students also have different supplementary materials
depending on their levels to develop writing skills.
The participants of the study are from two EIC 4 classes coded EIC4 – 0500 and EIC4 – 0200.
The two classes are chosen as intact groups for the study, in which, class EIC4 – 0500 (n=37)
is the experimental group and class EIC4 – 0200 (n=37) is the control group. The demographic
breakdown showed 33.8% male and 66.2% female participants. Both groups used the "English
File – Intermediate" textbook (units 6-10) and attended three 2.5-hour sessions weekly. The
experimental group practiced pronunciation using the SpeechAce platform, while the control
group followed the textbook-based approach.
Design of the Study
To address the research questions, the study utilized a mixed-method quasi-experimental
design. This approach, combining quantitative methods (pretests and post-tests) with qualitative
methods (questionnaires and interviews), allows for a comprehensive evaluation of

92
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

SpeechAce's impact on pronunciation competency. The experimental design, involving control

and experimental groups, establishes a causal link between SpeechAce use and pronunciation
improvement. Although convenience sampling was used due to practical constraints, the
approach remains effective in educational research. The combination of quantitative and
qualitative data provides a holistic understanding of both the efficacy of SpeechAce and the
students' perceptions of its use.
Data collection & analysis
The instruments used to collect data and assess the effectiveness of SpeechAce in enhancing
pronunciation competency among non-English majors are elaborated, as follows:
SpeechAce-Based Pronunciation Task
Participants from the experimental group will receive pronunciation training with SpeechAce.
To replicate the exact paper-based pronunciation task into SpeechAce is not possible. As a
result, the researcher made use of all available features in SpeechAce to maintain the same
characteristic between the SpeechAce-based pronunciation task and paper-based pronunciation
task. Fortunately, SpeechAce is designed to integrate with the LMS system Moodle, hence there
are more available features that can be adapted into SpeechAce to maintain its similarity with
the paper-based pronunciation task.
Pre-Test and Post-Test
Both groups completed a written pretest and post-test to assess their pronunciation competency.
The pretest comprised 37 questions from Test Your Pronunciation by Michael Vaughar (2002),
focusing on phonemes, stress, intonation, and phonemic transcription. The post-test, consisting
of 74 questions, was based on English Pronunciation in Use by Mark Hancock (2003). These
phonological elements are fundamental to pronunciation assessment, as they form the basis of
effective communication. The researcher graded both tests using the provided answer keys from
the textbooks, making an inter-rater unnecessary.
Questionnaires and Interviews
The experimental group completed a 23-item questionnaire on a five-point Likert scale to gauge
their attitudes towards pronunciation learning and using SpeechAce.
The study utilized standardized open-ended interviews with ten volunteers (2 males, 8 females)
conducted individually in a classroom setting and recorded with their consent. Each four- to
five-minute interview followed a consistent nine-question format. Participants, coded as S1 to
S10, answered questions categorized into three sections: their background and challenges in
learning pronunciation, perceptions of AI-based pronunciation training, and experiences with
SpeechAce.
Results/Findings
The Impact of SpeechAce on Students’ Pronunciation Competency
To assess the impact of SpeechAce on students' pronunciation competency, the study analyzed
data from the pretest and post-test scores of both the experimental and control groups.

93
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

Pretest
Table 1
Descriptive statistics of pretest scores
Group N Mean Standard Deviation
Control Group 30 30.93 5.95
Experimental
Group 36 31.47 4.30

Table 2
Independent Sample T-Test of pretests results
p- Sig. (2-
Pretest
value tailed)
Equal variances assumed 0.167 0.671
An Independent Sample T-Test was conducted to ensure that both groups started with
comparable pronunciation skills. As in table 1, the pretest mean scores were 30.93 for the
Control Group (CG) and 31.47 for the Experimental Group (EG). The slight difference in mean
scores (M = 30.93, SD = 5.95 for CG; M = 31.47, SD = 4.299 for EG) was evaluated for
statistical significance. The results of the Levene's test (Sig. = .167) in table 2 confirmed equal
variances, and the T-test showed a Sig. (2-tailed) value of .671, indicating no significant
difference between the groups. Thus, both groups were deemed equivalent in pronunciation
competency before the treatment.
Post-test
Table 3
Descriptive Statistics of Posttest Scores
Standard
Group N Mean
Deviation
Control Group 30 51.8 5.839
Experimental
36 58.03 4.687
Group

Table 4
Independent Sample T-Test of Posttests Results
p- Sig. (2-
Posttest
value tailed)
Equal variances assumed 0.257 0.00
As shown in table 3 and 4, the same analysis was performed on the post-test scores. The Control
Group had a mean score of 51.80 (SD = 5.839), while the Experimental Group scored
significantly higher, with a mean of 58.03 (SD = 4.687). The Levene's test again confirmed
equal variances (Sig. = .257). The T-test results revealed a Sig. (2-tailed) value of .00, indicating
a statistically significant difference between the post-test scores of the two groups. This

94
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

demonstrates that the Experimental Group, which used SpeechAce, showed a significantly
greater improvement in pronunciation competency than the Control Group.
Students’ Perceptions of Learning Pronunciation with SpeechAce
Results from questionnaires
The questionnaire was delivered to the students of the experimental group at the final session.
It is used to gauge students’ attitude and difficulties in pronunciation as well as their views
towards adapting SpeechAce into their pronunciation learning. Initially, there were 24 items in
the questionnaire, however, after the pilot test, one item was removed to satisfy the reliability
and validity of the questionnaire. The questionnaire then consists of 23 items, divided into three
sections.
The first section consists of 6 items to measure student’s attitudes toward pronunciation and 1
item to survey students’ preference in learning pronunciation. In the second section, students
express their difficulties in learning pronunciation through a set of 7 items. The final section
includes 8 items in multiple choice Likert scale form and one open-ended question. It surveys
students’ attitude towards using SpeechAce to learn English pronunciation. The questionnaire
was developed on a five-point Likert scale, spanning from (1) strongly disagree to (5) strongly
agree.
To ensure the questionnaire's reliability, Cronbach's Alpha was calculated for each section
using SPSS version 28. The results showed acceptable reliability levels with Cronbach's Alpha
values of 0.708, 0.723, and 0.755 for the three sections, respectively. These values indicate that
the questionnaire was both valid and dependable, providing a solid foundation for analysing
students' responses.
Students’ Perceptions and Difficulties in Learning Pronunciation
Figure 3
Students' perceptions towards pronunciation

95
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

The first section (figure 3) revealed that all students agree that having good English
pronunciation is essential for them, with 19 students (59.38%) strongly agree with this
statement and 13 students (40.63%) agree with this. Most students showed their desire in
learning pronunciation with half of the students surveyed found it interesting to learn
pronunciation, while only 2 students chose neutral. From the first two items’ results, students
have very serious attitude towards learning English pronunciation, not only did they find it is
an important component to acquire, but they also found interest in learning it. Most students
target native-like pronunciation with more than half of the participants agree with this item and
nearly half of them strongly agree with this, while only 2 students express a neutral stance and
none of the surveyed participants strongly disagreed or disagreed with this item. The result in
item 3 seems to synthesize with that of item 4, which ask participants to confirm whether they
think the optimal way to perfect English pronunciation is by speaking to native speakers. The
result shares similar pattern with item 3’s results, with nearly 90% of the participants agreed or
strongly disagreed, whereas no participants disagreed or strongly disagreed with this assertion.
In item 6, a considerable number of participants viewed it possible to achieve native-like
pronunciation. Only 1 student disagreed with this statement, meanwhile 8 of them took a neutral
stance and the rest of the group either agreed or strongly agreed. It seems that students believe
it is possible to speak English like a native speaker with proper practice and learning techniques.
However, item 5’s result has shown that students do not receive on-going pronunciation
learning, with 24 participants chose neutral when being asked to confirm if textbooks are their
main sources of learning pronunciation.
Figure 4
Students’ difficulties in learning pronunciation

The next section explored the challenges students faced. As shown in figure 4, over half of the
participants admitted they did not know how to learn pronunciation independently, with 17

96
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

agreeing and 3 strongly agreeing with this statement. Nearly half struggled with reading
phonemic transcriptions, and many found it difficult to find native speakers to practice with.
Ending and linking sounds were particularly troublesome, with a significant number of students
agreeing that these aspects were challenging. Despite these difficulties, most students disagreed
with the statement that teachers spend little time on pronunciation, and there were mixed views
on whether learning from different teachers’ accents caused confusion.
Students’ Perceptions of SpeechAce
Figure 5
Students' perceptions towards SpeechAce

Figure 5 revealed that students generally viewed SpeechAce positively. Nearly all found it easy
to use, and the majority agreed that it helped improve their pronunciation. SpeechAce also aided
students in reading phonemic transcriptions, with more than half acknowledging its
effectiveness in this area. The tool was particularly helpful in addressing errors related to ending
sounds, and students appreciated the detailed analysis of their pronunciation. Many students
reported that they could effectively imitate the model pronunciations provided by SpeechAce
and were able to learn independently using the tool.
The final open-ended question (“do you like learning pronunciation with SpeechAce? Why?
Why not?) revealed that 100% of the participants (n = 34) reported enjoying learning
pronunciation with SpeechAce. Notably, 76% (n = 26) expressed a strong interest in
incorporating SpeechAce into all their pronunciation lessons. The most frequently cited reasons
were that the tool is “interesting” (68%), “useful” (82%), and “easy to use” (74%). Additionally,
88% of the participants appreciated the detailed feedback provided by SpeechAce, which
helped them clearly identify specific areas of improvement. Overall, all participants (100%)
agreed that SpeechAce significantly contributed to enhancing their pronunciation skills.

97
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

Results from interviews

Students’ Background and Difficulties in Learning Pronunciation
When being asked about how they have been learning pronunciation previously, and whether
they find them effective, the students shared similar answers. Most participants shared that they
primarily learned pronunciation through formal instruction at school or English language
centers. One student (S5) noted, "I was lucky to have a native speaker teach our class for a
short course—it really helped me hear the correct sounds more clearly." Others mentioned
relying on digital platforms. As one student (S7) put it, "I often watch short clips on YouTube
or Facebook—especially the ones that focus on how to pronounce difficult words." Another
(S9) added, "Watching movies and listening to English songs has helped me a lot with
pronunciation, especially with stress and intonation." Additionally, several students mentioned
using mobile apps. For example, S4 said, "I use the Cake app almost every day because it gives
me short, useful dialogues to practice with." Similarly, S10 shared, "TFLAT is easy to use and
I like how it lets me listen and repeat after native speakers.". However, when it comes to the
effectiveness of their pronunciation learning process, most students expressed uncertainty about
the improvement of their pronunciation. S9 said that “it only has short-term effectiveness.
However, I can’t recall what I learned later on”. Similarly, S5 deemed her pronunciation
learning effective, but she does not think it has been improved. S3 stated that “the pronunciation
teaching methods at school may work for students with aptitude, but not for me”. In terms of
the difficulties they have in learning pronunciation, there were four main ones. Four students
reported that different methods of teaching pronunciation from different teachers confuse them.
“there were various teachers with different accents and pronunciations during my learning
process. As a result, each year we had to get used to different methods of pronunciation
teaching. This, consequently caused us great confusion.” – S2. This opinion is also shared by
S1, S4, and S3
Five out of ten interviewees said that they do not know if their pronunciation is correct since
no one helps them pinpoint their mistakes. S6 stated that “I don’t know how to pronounce the
word, so there should be something [like an app] or someone to correct my pronunciation”, or
S5 admitted that she often pronounced incorrectly, but she didn’t receive any feedback or
correction. S9 said that the app Cake she is currently using only grades her pronunciation on a
“good”, “not good enough” scale, and it doesn’t offer more thorough analysis of her
pronunciation. S10, a student who uses TFLAT, also shared a similar response.
Linking and ending sounds are also English pronunciation features which are troublesome to
the students. S1, S7, and S8 confessed that they have trouble in pronouncing and recognizing
these features in English pronunciation. S1 said she couldn’t link the sound together, while S7
said she was unable to recognize fast continuous linking speech. S8 stated that she couldn’t
manage to pronounce the weak sounds or ending sound /t/. Having no or little knowledge in
interpreting phonemic transcription is another difficulty reported by two of the interviewees,
S1 and S7.
It can be inferred that the pronunciation learning of the students took place in a relatively
spontaneous manner, without any fixed methods applied to their learning process. This, thus,

98
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

has caused confusion and difficulties in the way the students perceive their pronunciation
competency after some time of learning it.
Students’ Perceptions of Using AI to Learn Pronunciation
The data provides insights into their familiarity and perceptions regarding the use of AI in
learning pronunciation. While all ten interviewees acknowledged their awareness of integrating
AI in pronunciation training, their experiences and opinions varied, reflecting a complex
picture. For instance, participant S1 recognized the potential of AI to enhance learning by
making pronunciation adjustments more native-like. S4 highlighted the technical capabilities
of AI, such as phonemic transcription. However, there was a common sentiment among students
that current AI applications, including TFlat, Cake, Duolingo, and Elsa Speak, though useful,
were limited in their feedback specificity. As S8 noted, these applications often provide a
general percentage score of pronunciation correctness but lack detailed feedback on specific
sounds or words that need improvement. This general satisfaction with AI's basic
functionalities, coupled with a desire for more in-depth phonemic analysis, suggests a gap
between user expectations and current AI capabilities in pronunciation training. Such findings
echo the research by Chapelle (2001), who emphasized the need for AI language learning tools
to provide more granular feedback to be genuinely effective. Therefore, while students are
optimistic about the role of AI in pronunciation learning, there is a clear demand for advanced
features that offer detailed, personalized feedback to address specific areas of pronunciation
As reported by the participants, all of them have heard and used A.I. in practicing their
pronunciation at some point. However, they doubt that the A.I. they had used help them improve
their pronunciation competency. They also expected the type of A.I. which further analyzes
their utterance in phonemic level.
Students’ Perceptions Towards Learning Pronunciation with SpeechAce
After using SpeechAce for ten weeks, the students shared their thoughts on the tool. Generally,
they appreciated the detailed feedback provided by SpeechAce. S6 stated, "I like that SpeechAce
analyzes my pronunciation error in a very detailed manner, it points out what sound I got right,
what I need to improve on. Furthermore, I can listen and imitate the model utterance it offers."
S4 enjoyed the ability to compare his pronunciation with the AI model, feeling that this feature
significantly improved his skills. S3 highlighted that matching her pronunciation with the model
often required multiple attempts, which she found both challenging and motivating.
However, some students faced difficulties while using SpeechAce. Three students cited slow
internet connections as a hindrance, leading to delays or inaccurate feedback. S6 and S1 wished
for an option to slow down the speech rate of the model utterances, with S1 noting, "I know that
machine [AI] doesn’t guarantee 100% accuracy, but I wish there would be a function which
allows you to slow the speech rate down so we can hear the utterance more clearly." S2 found
the process of accessing SpeechAce through the MLearning system inconvenient, while S7 was
frustrated by inconsistent scores. Despite these issues, S4 and S8 reported no significant
difficulties, apart from needing a quiet environment for practice.
When asked to compare SpeechAce with a real teacher, opinions were mixed. Four students
felt that SpeechAce was comparable to a real teacher, with S9 noting, "SpeechAce is relatively

99
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

good, and it pronounces quite similarly to real teachers." However, others preferred learning
from real teachers. S1 argued that real teachers help improve not only pronunciation but also
confidence and communication skills, stating, "Frankly, I still prefer learning with a real
teacher because we can genuinely express ourselves. As a result, when we meet foreigners, we
don’t have the feeling of shyness. AI, however, merely helps us improve our pronunciation, not
our mental confidence. That’s why I think learning with real teachers is better."
Overall, all students agreed that their pronunciation had improved after using SpeechAce. S4
said, "I felt that my pronunciation has clearly improved since using SpeechAce." S5 rated her
improvement from a 5 or 6 before using SpeechAce to a 7 or 8 afterward. The detailed analysis
and ability to retry pronunciation exercises were cited as key factors in their progress. S1
summed it up by saying, "SpeechAce has helped me understand my pronunciation mistakes and
the stress patterns, so I know where to stress and where not to."

Discussion
The data presented previously is further discussed and synthesized with previous studies
mentioned in literature review section. It aims to see how thoroughly the collected data answers
the two research questions. Thus, it is divided into two parts. The first part is based on the
collected data from pretest and posttest to discuss students’ improvement on pronunciation
competency after using SpeechAce. The second part combines data collected from
questionnaire and interviews with the experimental group to conclude on students’ perceptions
towards using SpeechAce to practice their pronunciation.
The Impact of SpeechAce on Students’ Pronunciation Competency
The pretest results show that EG's mean score was slightly higher than that of CG’s mean score
with 31.47 and 30.93 respectively (table 4.1). However, the Sig. (2-tailed) value obtained from
the independent samples test was .671 (> .05) (table 4.2). Thus, it can be concluded that the
differences between the means of CG and EG were relatively insignificant. After eleven weeks
of applying SpeechAce to the EG, the posttest results show that EG’s mean score is now
relatively higher than CG’s mean score, at 58.03 compared to 51.80 (table 4.3). Moreover, the
2-tailed Sig. value was found to be .00 (table 4.4), which is below the threshold of .05. This
result indicates a statistically significant difference between the posttest means of the control
group (CG) and the experimental group (EG). Therefore, it can be confirmed that the
pronunciation competency of both groups improved after the treatment, with the performance
of the experimental group notably surpassing that of the control group. This finding parallels
with the results from the three studies previously mentioned in the literature review section by
Aiello and Mongibello (2019), Moxon (2021), and Lidia (2021). However, Aiello and
Mongibello acknowledged the lack of control group in their study, and Moxon He admitted the
study's restrictions stemming from the absence of male students and limitations in terms of
time. This study, which was done with both control group and experimental group during a
process of 10 weeks, has bridged the gaps mentioned by previous authors. All these factors
validated the impact of improving students’ pronunciation competency through the utilizing of
SpeechAce.

100
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

Students’ Perceptions of Learning Pronunciation with SpeechAce

Prior to the study, it was anticipated that students might undervalue the importance of learning
pronunciation, as some exhibited unintelligible pronunciation and showed limited effort in
improving this skill. This perception was further influenced by the researcher’s teaching
experience, which involved only a few isolated sessions on pronunciation, with most instruction
occurring spontaneously and informally. However, the data collected from the questionnaire
revealed unexpected results. The responses indicated that all participants held a serious attitude
toward learning pronunciation. Specifically, all students either agreed or strongly agreed with
the statement that 'it is very important to possess good pronunciation,' and the majority
expressed interest in learning pronunciation. This positive attitude was contrary to the
researcher’s initial expectations. Additionally, most students displayed optimism about their
ability to achieve a native-like accent, aligning with the findings of Dao (2018).
Unfortunately, such enthusiasm and positivity in learning pronunciation of the students were
hindered by the difficulties they had to face. In fact, many students’ difficulties matched with
previous points mentioned in the literature review chapter. Specifically, nearly two-thirds of
the surveyed participants admitted that Students reported a lack of knowledge about effective
methods for independently improving their pronunciation, despite many disagreeing or strongly
disagreeing with the statement that 'teachers spend little time teaching pronunciation.' This
discrepancy may be attributed to the tendency of many instructors to approach pronunciation
instruction in an unstructured manner, often addressing issues reactively based on immediate
classroom needs rather than through preplanned methods.(Nguyen & Newton, 2020). The
challenge of lacking effective methods for learning pronunciation was also highlighted by four
participants during the interviews (S1, S2, S3, S4).
There were various teachers with different accents and pronunciation during my learning
process. As a result, each year we had to get used to different methods of pronunciation teaching.
This, consequently caused us great confusion.” – S2. This opinion is also shared by S1, S4, and
S3

In the previous chapter, the researcher assumed that most Vietnamese English learners had
difficulty recognizing and pronouncing ending and linking sounds. Certain English phonetic
sounds, including /t/, /ʧ/, /ʤ/, /ʒ/, /s/, /θ/, and /ʃ/, were identified as particularly challenging
(Tam, 2005). These findings are supported by the questionnaire results: 72% of students agreed
or strongly agreed that they struggled with ending and linking sounds, while 18% remained
neutral and 10% disagreed. Similarly, interview responses from three participants (S1, S7, S8)
echoed these concerns. In addition, a significant portion of the students (65%) reported
unfamiliarity or lack of confidence in using phonemic transcriptions for learning pronunciation,
while 22% expressed neutrality and 13% felt confident using them.
With the above attitude and difficulties in learning pronunciation, the students received
SpeechAce very positively. First, 31 out of 32 surveyed participants said that SpeechAce was
easy to use and 25 of them believed they could learn pronunciation effectively by themselves,
the rest took a neutral stance. This may solve the problem of times constraint in teaching and
learning pronunciation in class. Students’ problem in understanding phonemic transcription, to

101
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

an extent, might be solved with SpeechAce, with nearly two thirds of the participants believing
that they could read phonemic transcription, while 11 others chose neutral for this point. A
similar pattern is observed in the potential of SpeechAce to assist students in addressing issues
with missing ending sounds. One specific aspect of SpeechAce that students found particularly
engaging was its ability to provide detailed feedback on their pronunciation. Almost all
students, except one chose neutral, stated that they knew exactly which sound they pronounce
incorrectly and 20 of them said they could recreate the sound by listening to and imitating the
model utterance. This assertion was later developed in the interview, in which all ten
interviewees reaffirmed this good point of SpeechAce. In the literature review, a number of
authors mentioned the effectiveness of corrective feedback in teaching and learning
pronunciation (Coupe, 2015; Ellis, 2009; Kim & Han, 2007; Lee et al., 2015; Long, 2015;
Lyster et al., 2013; Saito & Lyster, 2012). Undoubtedly, SpeechAce has offered the appropriate
form of corrective feedback that learners require to enhance their pronunciation learning with
greater effectiveness. Finally, most students admitted that their pronunciation had improved
after using SpeechAce, all participants in the interview also supported this view and half of the
surveyed participants expressed that SpeechAce could help them pronounce like native
speakers.
It seems that SpeechAce has broken down all barriers that stop students from learning
pronunciation effectively. Also, it has helped them improve their pronunciation in certain areas.
However, it should be noted that there exist some uncertainty amongs students’ answers. Thus,
it would never be a good idea to rely solely on SpeechAce to teach pronunciation, as many of
the interviewed students reported that they still preferred real teachers to teach them
pronunciation because it is not only the proper pronunciation they are looking for, they also
wish to talk with real people to boost their confidence in communication. All in all, it would be
wise to use SpeechAce only as an effective tool to help teachers bridge the gaps in teaching
pronunciation to students, and the idea of A.I that can completely replace a human teacher has
yet to come.

Conclusion
The study aims to investigate the impact of SpeechAce in developing pronunciation
competency for non-English-majored students at a private university. It seeks to answer the two
questions: “to what extent does SpeechAce improve students’ pronunciation competency?” and
“what are the students’ perceptions toward learning pronunciation with SpeechAce?”. A mixed
method quasi-experimental design with one control group (CG) and one experimental group
(EG) was applied. The results from the pretest and posttest have shown that students’
pronunciation competency of the EG was improved compared to that of CG and thus, confirmed
the effectiveness of using SpeechAce to improve pronunciation. The data obtained from the
questionnaire and interviews have also revealed that the majority of students received
SpeechAce positively although there lie some uncertainties which calls for teachers’ careful
consideration and preparation when using SpeechAce to teach pronunciation.

102
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

For educators, the study underscores the importance of recognizing students' enthusiasm for
learning pronunciation, aligning with Gilbert's (2008) emphasis on leveraging students'
willingness to improve. Despite the challenges, such as the time-consuming nature of
pronunciation instruction, teachers have a critical role in making pronunciation learning
meaningful and engaging for students. Derwing and Munro (2005) highlight the necessity for
teachers to not only instruct but also model accurate pronunciation to facilitate students'
acquisition. Additionally, teachers must focus on correcting common pronunciation errors
specific to Vietnamese English learners, as indicated by Jenkins (2000). Incorporating
phonemic transcription into teaching, as suggested by Celce-Murcia et al. (2010), is crucial, as
it has been proven in this study to significantly benefit students’ learning process. Above all,
embracing AI technology, such as SpeechAce, and integrating it into teaching practices can
enhance the effectiveness of both learning and teaching, as recommended by Chapelle (2001).
For students, SpeechAce provides a valuable resource for independent pronunciation practice.
However, effective use of this tool requires students to become proficient in interpreting
phonemic transcriptions, a skill essential for pronunciation learning, as noted by Levis (2005).
This also necessitates a degree of self-discipline for consistent practice.
For curriculum designers, the positive impact of SpeechAce underscores the need for
incorporating AI technology into English language education, supporting Pennington's (1996)
argument for technology integration in language learning. Additionally, formal pronunciation
training for teachers can enhance their confidence and competence, thereby benefiting students,
as emphasized by Derwing and Munro (2015). Institutions could integrate SpeechAce into
curricula through blended learning models, where students receive immediate, individualized
pronunciation feedback alongside traditional instruction
While the study achieved its objectives, certain limitations must be acknowledged. The study's
sample size was constrained by practical considerations, potentially limiting the generalizability
of the findings to broader populations. Additionally, the study focused on a specific
demographic, and the impact of SpeechAce on learners from diverse linguistic backgrounds
might differ. The participants' familiarity with technology, particularly ASR-based tools like
SpeechAce, could have influenced their engagement and outcomes, and the transformation of
textbook exercises into the SpeechAce format was not always accurate, potentially affecting
the validity of the activities. Furthermore, the pretest and post-test design did not involve oral
pronunciation tasks, which could have provided a more reliable assessment of pronunciation
competency. Finally, the lack of inter raters is another factor that does not guarantee a high
level the reliability and validity of the study.
Future research should involve larger sample sizes to assess the effectiveness of ASR
technology in teaching pronunciation. Pretests and post-tests should be designed to evaluate all
components of English pronunciation and include oral tasks to accurately gauge competency.
The inclusion of inter-raters in grading will enhance the objectivity of results. Additionally,
further studies should explore the most effective pedagogical strategies for integrating ASR
into language instruction, considering diverse learner populations and investigating the impact
of ASR on other language skills, such as listening comprehension and speaking fluency.

103
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

References
Bajorek, J. (2017). L2 pronunciation tools: The unrealized potential of prominent computer-
assisted language learning software. Issues and Trends in Educational Technology, 5(1).
[Link]
Celce-Murcia, M., Brinton, D., & Goodwin, J. (2010). Teaching Pronunciation: A Reference
for Teachers of English to Speakers of Other Languages. Cambridge: Cambridge
University Press.
Derwing, T., & Munro, M. (2005). Second language accent and pronunciation teaching: A
research-based approach. TESOL Quarterly, 39(3), 379-397.
Derwing, T. M., & Rossiter, M. J. (2002). ESL learners’ perceptions of their pronunciation
needs and strategies. System, 30(2). [Link]
Fraser, H. (2000). Coordinating improvements in pronunciation teaching for adult learners of
English as a second language. Canberra: DETYA (ANTA Innovative Project). Retrieved
from [Link]
Gilakjani, A. P. (2012). The significance of pronunciation in English language teaching.
English Language Teaching, 5(4), 96-107. [Link]
Gorjian, B., Hayati, A., & Pourkhoni, P. (2013). Using Praat software in teaching prosodic
features to EFL learners. Procedia - Social and Behavioral Sciences, 84.
[Link]
Ha, T. T. X., & Bao, N. K. V. (2023). Common errors in pronunciation of non-English majored
students at the university of Transport and Communication Ho Chi Minh Campus. South
Asian Research Journal of Arts, Language and Literature, 5(03), 64-70.
[Link]
Hassim, H., Kassim, H., Kassim, A., & Kassim, M. (2023). Exploring the use of artificial
intelligence-based technology to enhance creativity in ESL speaking classroom. The
Proceedings of 7th IEEE Congress on Information Science and Technology (CiSt).
[Link]
Hirai, A., Kovalyova, A. (2023). Using speech-to-text applications for assessing English
language learners’ pronunciation: A comparison with human raters. In Suárez, MdM., El-
Henawy, W.M. (eds) Optimizing Online English Language Learning and Teaching.
English Language Education, vol 31, 337-355. Springer, Cham.
[Link]
Horwitz, E. (2001). Language anxiety and achievement. Annual Review of Applied Linguistics,
21, 112-126. [Link]
Jiang, M. Y. C., Jong, M. S. Y., Lau, W. W. F., Chai, C. S., & Wu, N. (2023). Effects of
automatic speech recognition technology on EFL learners’ willingness to communicate
and interactional features. Educational Technology and Society, 26(3), 37-52.
[Link]
Kholis, A. (2021). Elsa speak app: Automatic speech recognition (ASR) for supplementing
English pronunciation skills. Pedagogy: Journal of English Language Teaching, 9(1), 1-
14. [Link]

104
CALL-EJ Computer-Assisted Language Learning Electronic Journal Vol. 26; No. 3; 2025

Kumar, L. A., & Renuka, D. K. (2023). Deep Learning Approach for Natural Language
Processing, Speech, and Computer Vision. CRC Press.
[Link]
Levis, J. (2007). Computer technology in teaching and researching pronunciation. Annual
Review of Applied Linguistics, 27, 184-202. [Link]
Levis, J. M. (2005). Changing contexts and shifting paradigms in pronunciation teaching.
TESOL Quarterly, 39(3), 369-377. [Link]
Li, J., Deng, L., Haeb-Umbach, R., & Gong, Y. (2015). Robust Automatic Speech Recognition:
A Bridge to Practical Applications. eBook ISBN: 9780128026168
Loc, N. T., & Newton, J. (2020). Pronunciation teaching in Tertiary EFL classes: Vietnamese
teachers' Beliefs and Practices. The Electronic Journal for English as a Second Language,
24(1), 1-20.
Moxon, S. (2024) A review of the website [Link], Computer-Assisted Language
Learning Electronic Journal. Available at:
[Link] 09 May 2025).
Moxon, S. (2021). Exploring the effects of automated pronunciation evaluation on L2 students
in Thailand. IAFOR Journal of Education, 9(3), 41-56. [Link]
Nguyen, H. N. (2023). Teaching English sounds to Vietnamese secondary school students:
From theories to applications using learner-centered methods. International Journal of
Language Instruction, 2(1), 16-29. [Link]
Vu, N. N. (2024). Computational Linguistics: From Theory to Practice. Hồ Chí Minh: NXB
ĐHSP [Link].
Nguyen, T. (2017). Vietnam’s national foreign language 2020 project after 9 years: A difficult
stage. The Asian Conference on Education and International Development.
Sholekhah, M. F., & Fakhrurriana, R. (2023). The use of ELSA Speak as a mobile-assisted
language learning (MALL) towards EFL students’ pronunciation. JELITA: Journal of
Education, Language Innovation, and Applied Linguistics, 2(2), 93-100.
[Link]
Spring, R. and Tabuchi, R. (2024) The role of ASR training in EFL pronunciation improvement:
An in-depth look at the impact of treatment length and guided practice on specific
pronunciation points, Computer-Assisted Language Learning Electronic Journal.
Available at: [Link] (Accessed: 09 May
2025).
Trofimovich, P., & Isaacs, T. (2012). Disentangling accent from comprehensibility.
Bilingualism, 15(4), 905-916. [Link]
Vu, N. N. (2016a). An investigation of Vietnamese students’ learning styles in online language
learning. Journal of Science, Ho Chi Minh University of Education, 79(1), 16-24.
Vu, N. N. (2016b). Mobile learning in language teaching context of Vietnam: An evaluation of
students’ readiness. Journal of Science, HCMC University of Education, 7(85), 16-27.
Retrieved from [Link]

105
[Link] Nguyen Ngoc Vu, Vo Thiet Thach, Tran Vu Diem Thuy Vol. 26; No. 3; 2025

Biodata
Assoc. Prof. Dr. Nguyen, Ngoc Vu is an Associate Professor of Linguistics at Ho Chi Minh City
University of Foreign Languages - Information Technology (HUFLIT), Vietnam and Chairman
of STESOL, a TESOL organization setup by Association of Vietnamese Universities and
Colleges. His research interests include Computational Linguistics, Teaching English to
Speakers of Other Languages (TESOL) and Computer-Assisted Language Learning (CALL).
Mr. Vo, Thiet Thach holds a Master of Arts in TESOL and is an experienced English lecturer
at Ho Chi Minh City University of Foreign Languages - Information Technology, Vietnam. With
over 10 years of university-level teaching, his research interests include English for Specific
Purposes (ESP), Computer-Assisted Language Learning (CALL), and Linguistics.
Dr. Tran, Vu Diem Thuy is currently working as Head of Department of Applied Linguistics at
the Faculty of Foreign Languages, Ho Chi Minh City University of Foreign Languages -
Information Technology, Vietnam. She has been teaching English for over 22 years at the
university level. Her research interests include ESP, CALL/MALL, Linguistics, Computational
Linguistics, and Teacher Education.

106