Integrating AI into Second Language Pronunciation Teaching and Learning

Seungeun Lee

Artificial Intelligence (AI) has attracted significant global attention regarding its potential to influence daily life and even transform educational practices. In particular, AI’s capacity to generate teaching materials has raised expectations for its usefulness in areas such as pronunciation, which has often received less focus in second language instruction. In this article, I share reflections and practical recommendations based on my experience teaching K-12 learners of Korean as an additional language. These insights highlight the use of generative AI to support language teaching and learning, specifically addressing the challenges I encountered and the strategies I employed to overcome them.

To structure my sessions, I adopted the five-step framework for pronunciation instruction proposed by Celce-Murcia et al. (2010), which offers a systematic approach to teaching pronunciation. The steps and their related activities are shown below:

Step 1 - Description & Analysis: Illustrating how the feature is produced and when it occurs within spoken discourse

Step 2 - Listening Discrimination: Focused listening practice with feedback on learners’ ability to correctly discriminate the feature

Step 3 - Controlled Practice: Oral reading of minimal-pair sentences, short dialogues, etc., with special attention paid to the highlighted feature in order to raise learner consciousness

Step 4 - Guided Practice: Structured communication exercises, such as information-gap activities or cued dialogues, that enable the learner to monitor the specified feature

Step 5 - Communicative Practice: Less structured, fluency-building activities (e.g., role play and problem solving) that require the learner to attend to both form and content of utterances

Throughout a year of teaching Korean phonemes and connected speech features, regardless of which segmental or suprasegmental features were taught, all sessions followed the same sequence, from Step 1 (Description & Analysis) to Step 5 (Communicative Practice). Although the teaching targets and instructional activities varied in each session, I identified three general tips based on how I overcame difficulties with the assistance of AI while teaching according to the five steps.

1) From Step 1 (Description & Analysis) to Step 2 (Listening Discrimination)

TIP 1. Consider using generative AI to find example words that contain target sound features. However, if the AI outputs nonwords or incorrect responses, teachers can creatively find ways to transform these error-like outputs into diverse learning opportunities.

In these stages, learners begin to gain explicit knowledge of target features and apply them in their sound perception process. It is recommended to prepare different and varied word examples. This is where I found generative AI useful. For instance, I input prompts such as “Create example words for Korean learners to practice ㄴ-insertion, for example, like the word 담요 pronounced as 담뇨. Provide multiple noun examples.” This way, I could get many examples in a very short time, which thankfully saved me from time constraints and a lack of creativity in finding example words. However, using generative AI is not without caveats. First, I needed to double-check if the words were deemed appropriate given the academic and cognitive level of my adolescent students. Sometimes, the generative AI provided words that seem to be used infrequently in daily lives or to be conceptually unfamiliar to students, such as 지향성 (intentionality). For these cases, I had to exclude the words for the smooth flow of sessions. Moreover, often generative AI can even output nonwords. One way you can consider is to think about ways to employ those responses in a new manner, rather than ignoring them as an error. These nonwords can be used by designing a pedagogical activity to help learners find a correctly spelt word between two words that are phonologically similar but orthographically different. I asked students to pick the correct word after listening to me read the names of several foods. Through this activity, I could raise students’ awareness that pronunciation and orthography may not correspond in some conditions.

2) Step 3 (Controlled Practice)

TIP 2: Use generative AI to customize materials and contextualize them based on learners’ current learning situations or contexts.

This stage is to help learners practice words or sentences to pronounce target sound features within a larger discourse context. In order to do this, using sentence-level examples is encouraged. Since I was using a textbook that was initially designed for K-12 learners speaking Korean as a mother tongue, sometimes dialogues or reading passages were deemed pedagogically not fit with my students. Given this situation, I used generative AI to customize and edit the existing passages whose context seemed unfamiliar to my students. For instance, while keeping original structures and target words of the passages, I changed the initial passage about the themes of Joseon and the Korean Peninsula to one about other topics that intrigued my students more (e.g., American celebrities and Hawaii).

3) Step 4 (Guided Practice) to Step 5 (Communicative Practice)

TIP 3. When you are drained of creativity in brainstorming activities or materials, generative AI might spark new ideas. However, please note that teachers should be the one who makes the final decision of whether the materials or activities generated by AI could be the most appropriate and beneficial for your students. 

Now it is time to loosen teachers’ control but encourage students’ free production. In these stages, the activities are recommended to be less structured so that students can link their new knowledge to their existing language repertoire by applying it in their own utterances by themselves. However, it was never easy for me to think of different activities for every session that would pedagogically benefit students and facilitate their production of new sounds, while still being new enough to prevent students from getting bored with all the same patterns and activities. This is where I found generative AI helpful for pronunciation teachers to be more creative and make materials more interesting for students. I input prompts such as “tell me some games that let students practice the words below (담요 and 담뇨) while interacting with multiple different people. They must talk to a different partner each time, and they need to listen to each other’s utterances and communicate”. By using these prompts, I was able to brainstorm new ideas and find many useful interactive activities. However, sometimes you would see AI give you so many answers which seem to lack thought–a breadth of answers without depth. It is the teachers' responsibility to decide which activities would be the most pedagogically appropriate for their students and to always be willing to revise and build upon AI’s suggestions.

To conclude, chatting with generative AI when you get stuck in planning your session can be a good starting point or sometimes even a great breakthrough, saving your time and sparking new ideas. Nevertheless, teachers should be the major and final decision makers to find the most pedagogically useful activities throughout the process of designing their class. Although my three tips are based on my experience working with adolescent students, I hope the insights presented here will be both applicable and beneficial across a wide range of educational contexts for pronunciation instruction.

​Reference

Celce-Murcia, M., Brinton, D. M., Goodwin, J. M., & Griner, B. (2010). Teaching pronunciation: A course book and reference guide (2nd ed.). Cambridge University Press.

Author bio: 

Seungeun Lee is a PhD student in the Department of Second Language Studies at the University of Hawai‘i at Mānoa. She has taught English and Korean to second language learners in Hawai‘i and Korea.