LSTM-driven CLIL: Cybersecurity vocabulary learning with AI

This study presents the development of a custom dataset of L2 gap-fill exercises designed to enhance Long Short-Term Memory (LSTM) neural networks in CLIL (Content and Language Integrated Learning) settings for subject-specific courses. Targeting English for Special Purposes (ESP) vocabulary in cybersecurity, privacy, and data protection, the model addresses the dual challenge of domain-specific context mastery and language practice through structured neural network training. The custom dataset of gap-fill exercises for this LSTM model enables simultaneous prediction of missing words and semantic classification, offering learners contextualized language training that is a core requirement of CLIL methodology. Experimental results validate the model’s efficacy, demonstrating its potential as an adaptive support tool for CLIL-based education. This framework establishes a novel synergy between AI-enhanced language learning and subject-specific instruction, providing a scalable template for integrating neural networks into CLIL pedagogy.