GENERATING PERSONALISED FLASHCARDS AND EXAMPLES TO HELP LEARN VOCABULARY USING AI

Зулуфова Милена Болебековна

Студентка Таразского университета имени М.Х. Дулати,

Факультет «Филологии и гуманитарных наук», г.Тараз

If you ask an adult who has learnt a foreign language on their own what is most frustrating about learning and memorising vocabulary, the answer is almost always the same: flashcards. To be more precise, it is not the format itself, but the endless cycle of ‘word – translation – next flashcard’.

A word taken out of context is difficult to remember. Learning within a meaningful context, rather than in isolation, leads to significantly higher rates of word retention; according to some reviews, the difference is as much as 30 per cent. Context provides the brain with additional semantic and grammatical cues, which make it easier to retrieve the word from memory later on. Kohnke et al. highlight that generative AI’s capability to tailor learning materials to individual learner profiles significantly boosts intrinsic motivation, making language acquisition more meaningful for adult professionals[1].

The problem is that the standard examples in textbooks are, as a rule, extremely neutral: ‘The cat is on the table’, ‘Der Mann geht ins Büro’. Formally, there is a context, but in essence it belongs to no one; it has no connection to any particular student. And for an adult with limited time and motivation that isn’t always consistent, it is difficult to connect emotionally with a sentence about an abstract man walking into an abstract office.

Furthermore, there are factors that indicate the degree to which a word has been learnt. To fully master a word, collocation or phrase, one needs to understand its synonyms and antonyms; if it is a verb, its prepositions and verb government; how the word is spelled and pronounced; in what context it is used; and so on.

So what is the core essence of this approach and the use of AI in learning new words? Firstly, to eliminate neutrality. Instead of generating examples at random, examples are created specifically for each individual student, taking into account their real interests, hobbies, profession, and everything the teacher knows about them. For example, if a student is studying law and learning specialist vocabulary, instead of the dry ‘the contract was signed yesterday’, the model might suggest an example based on an analysis of a court case or a plot from a TV series about lawyers. If a student is keen on football, new words are woven into the context of the transfer window, referees’ decisions and offside rules. The aim is simple: the word should stick not only because of its grammatical structure, but also because it relates to something the learner finds genuinely familiar and interesting.

This is the key difference from the usual ‘contextual examples’ that are already widely used: we are talking about hyper-contextual, personalised examples tailored to a specific student, rather than to an average ‘typical language learner’.

This approach can be used:

1. Gathering information on students’ interests. Ask a few questions at the start of the course: what the student does, what they watch, and what they enjoy doing in their spare time. This is the usual introductory chat that many teachers conduct anyway.

2. Compiling a list of target words – the vocabulary to be covered at this stage: legal terms, sports vocabulary, medical terms, etc.

3. Generating sentences using a template that explicitly specifies a word and the student’s area of interest. For example: ‘Create an example using verb x in context y’.

4. Assessment via a mini-test – not just a flashcard with a ready-made sentence, but a question about the meaning of the word in that specific context, or a request to complete a similar sentence independently.

Personalisation works not only at the level of ‘interesting/uninteresting’. It also reduces cognitive load. When the situation in the example is already familiar to the student in terms of meaning, the brain does not need to expend resources on understanding the context, and all energy is directed towards the word itself. Research into personalised vocabulary learning apps is based on this principle: by reducing cognitive load and increasing engagement, personalised learning pathways ensure more robust long-term retention compared to generic flashcards.

Methodologists note that if, after introducing a new word, you ask a student a personalised question specifically related to that word, this helps to reinforce the vocabulary and, at the same time, enhances the atmosphere of the lesson, as people generally respond better to stories, especially their own. Furthermore, as well as generating examples, context and so on, AI can create an online test using these words, suggest other exercises or compose a complete text. In addition, an AI bot can generate a format that can be integrated into your favourite learning platforms, allowing you to practise with flashcards there. Alternatively, for those who prefer traditional and classical learning methods, you can ask it to generate flashcards for printing.

Personalisation via AI serves two purposes. On the one hand, it enhances memorisation: context generally works better than isolated words, and personal context is more effective than neutral context. On the other hand, it boosts motivation: material linked to the student’s real interests is better received than material about an abstract man and his office.

However, it is important not to overdo it, as an artificially contrived or absurd context can have the opposite effect. Therefore, this approach requires manual checking by the teacher, as the generated examples must be grammatically correct and stylistically appropriate, rather than merely formally ‘on topic’. The model, like any tool of this kind, sometimes produces strange results even when attempting to be personalised.

REFERENCES:

1.Kohnke, L., Moorhouse, B. L., & Zou, D. (2024). Promoting autonomous vocabulary acquisition through generative AI: Tailoring materials for adult ESP learners. TESOL Quarterly, 58(3), 745–760.