Video games have been steadily increasing in popularity since the early 1950s, with the creation of new platforms and more complex systems. By using immersive learning, where users can practice skills in virtual environments, this study explores the use of video games as a teaching tool for languages where no native speakers remain such as Latin. The motivation for this work stems from the need for more effective, engaging methods to learn ancient languages that lack natural conversational contexts. Traditional memorization-based methods are often limited in sustaining learner motivation, while immersive approaches can replicate real, communicative environments. Natural Language Processing (NLP) models such as BERT can be used to automate the process of grammar correction to improve a user’s vocabulary and grammar when interacting with non-player characters. We fine-tuned BERT on synthetically generated Latin error-correction data and integrated it into the game environment to provide realtime grammatical feedback to learners. Our results showed that model achieved high accuracy in detecting verb and noun errors, though its performance on adjective errors was less robust. Overall, the findings highlight the feasibility and promise of combining feedback from state-of-the-art large-language models (LLM) with game-based immersive learning to enhance Latin acquisition.
Zelek et al. (Thu,) studied this question.