Key AI technologies in language learning

Language lecturers are at the forefront of exploring which AI tools are effective, suitable and available, as well as of identifying practical guidelines for the ethical use of these tools by both students and teachers. However, all of these technologies and terms are still new to most of us.

As a language teacher seeking to better understand them and their usage, I have compiled a brief list outlining the technologies currently in use, accompanied by short descriptions and an overview of their benefits as reported in previous research.

AI in language learning and teaching

Based on some analyses AI supports teachers in their planning, implementing and assessment in their teachings (Celik et al. 2022). Additionally, the benefits of using AI in education for teaching, learning, assessing, evaluating, adapting, personalizing has been confirmed by research on many levels in education (Chiu et al. 2023; Yang & Kyun 2022).

The most commonly used AI technologies and applications in language learning can be categorized in seven interlinked groups: natural language processing (NLP), data-driven learning (DDL), automated writing evaluation (AWE), computerized dynamic assessment (CDA), intelligent tutoring systems (ITSs), automatic speech recognition (ASR), and chatbots. (Son et al. 2023.)

Below, I will present a selection of these technologies from a language learning perspective.

Natural language processing

The work on machine translation in the late 1940’s marked the beginning of the development process of today’s form of NLP (Khurana et al. 2023). Natural Language Processing (NLP) is a relatively new field with the aim of enabling a conversation between humans and computers (Jurafsky & Martin 2009); Khurana et al. 2023). The use of NLP can vary and in the context of language learning be applied in translating, extracting information, summarizing text, and question answering (Khurana et al. 2023). NLP and other techniques such as machine translation, user modeling and others are of the most used techniques in Language learning systems (Gamper & Knapp 2002).

In general, NLP technologies are used in other systems such as Automated writing evaluation AWE and automated essay scoring (AES) and others showing their positive effect on evaluation and scoring (Gamper & Knapp 2002).

From a language teacher’s perspective, one of the most evident applications of NLP is in translation. I still remember when translating phrases between English and Macedonian often produced funny and inaccurate results. For example, the English phrase ‘on the other side’ could be translated into Macedonian as ‘на другата страна’ (na drugata strana), which, when translated back into English, will be ‘on the other page.’ This illustrates how earlier systems lacked natural language processing (NLP) capabilities. Nowadays, however, NLP is widely applied, especially in translation, leading to much more accurate and context-sensitive results.

Data-driven learning (DDV)

The definition of Data-driven learning is still open for discussion and debate. DDL means using a corpora, software and technology to learn a language. In many instances corpora can mean many different things in the language learning context. Corpora represent a collection of authentic texts that are used by learners. Learners learn from investigating and communicating with technology to get an answer or understanding of the language issue. (Boulton 2016)

Boulton (2021) has stated that despite the criticism it has received, DDL has a positive effect on language learning. The key point for future improvement for normalisation of corpora use is teachers training and their active involvement, and technology advancement for language learning (Pérez-Paredes 2022).

As language teachers, we would need to gather a corpus of authentic data such as materials and news articles from various sources — which can be time-consuming. Students can then search the corpus for common phrases and patterns using real examples. Tools like SKELL (Sketch engine for language learning) are specifically designed to help learners discover word collocations and usage examples efficiently.

Automated writing evaluation (AWE)

With the advancement of AI technology, automated writing evaluation software has been developed and is used by students to receive holistic feedback on their writing (Liu 2024). AWE is software that uses sophisticated language algorithms to provide feedback and analytical scores. There are different available systems, and they all have similar or different algorithms. Most AWE software is based on natural language processing mechanisms, meaning their ability to analyse input text largely depends on the size of the selected corpora. (Roscoe et al. 2017.)

Liu’s (2024) comprehensive review states that AWE is widely used, and the effects are positive towards improving the learning outcomes. Fu et al. (2024) have reported that based on their research 53 % had a positive outcome, which means that the AWE systems have limitations and future improvements are necessary.

Computerized dynamic assessment (CDA)

Dynamic assessment (DA) is an approach in evaluation with guiding principles based on the Vygotsky’s sociocultural theory (Poehner et al. 2015). During DA the assessor or mediator provides support to the learner when assessing their performance to gain more vivid understanding of the learners’ progress and developed abilities (Kargar Behbahani & Karimpour 2024). Computerized DA emerged from computerized testing, and it combines instruction and assessment by offering digital support to learners. CDA focuses on identifying and addressing learning challenges through technology, beyond traditional tests to create a more supportive, interactive, and educational assessment environment.(Kargar Behbahani & Karimpour 2024; Poehner et al. 2015).

Using CDA in learning languages brings success in listening and reading not only for students, but it also helps teachers to give more individualised support for students. CDA offers flexibility in location and time for taking tests and other learning activities. Students learning while using CDA performe more efficiently than those who use conventional teaching and learning methods. (Son et al. 2023.)

Currently, with the use of AI, language teachers can create a wide variety of activities, such as word matching, sentence completion, and grammar exercises. Moodle now offers options for creating AI-powered quizzes. Unlike traditional quizzes, CDA allows for scaffolding that was previously provided only by a teacher. Scaffolding can now be incorporated automatically, and feedback can be provided in real time to guide learners. Also, CDA assesses and offers corrective feedback dynamically for spoken language. Among these tools, Duolingo remains one of the most familiar.

Language teachers are at the forefront of adopting AI technologies: institutional support remains crucial

It is evident that AI is already used in education and holds significant potential to enhance both teaching and learning. Yet, when the goal of a country or institution is to enhance AI literacy and improve AI pedagogy, there is a clear need for teacher support, guidance, and further research.

The main focus in research and publicans like UNESCO (Miao & Cukurova 2024) AI Competency Framework for Teachers, is in gaining skills, building teachers ‘competencies to use AI tools within their pedagogical methods to support the students learning, socialisation, social caring by developing their practices into inclusive and innovative. Celic (2023) and Law (2024) both emphasise the necessity of further research on the ethical use of generative AI in education, as well as on critical issues of data privacy and security.

Additionally, both teachers and students should be actively involved in the development of AI tools and provided with training on their ethical and responsible use.

Picture: Haaga-Helia