The advent of pretrained language models (PLMs) has revolutionized the field of natural language processing (NLP). Trained on extensive datasets, PLMs serve as foundational models for a myriad of NLP tasks. Over the past decade, various types of PLMs have emerged, ranging from static to contextual language models and from masked to autoregressive models, each offering unique ways of utilization and interaction.

This thesis explores these diverse PLM paradigms—embeddings, fine-tuning, prompt-based learning, and in-context learning—within the educational domain, where generative AI and PLMs are poised to effect significant transformation. By examining the application of these paradigms in educational settings, this work aims to highlight the potential of PLMs to enhance teaching and learning processes, thereby contributing to the ongoing evolution of technology-enhanced learning (TEL).

The exploration contributes to the educational domain by expanding knowledge on PLMs and demonstrating practical methods for integrating generative AI into educational practice. The studies presented in this thesis have concretely improved upon the state-of-the-art approaches for various tasks, enhancing current methods and theories in both the TEL and NLP domains. These advancements contribute to the evolution and improvement of both fields, offering innovative applications of AI in education while advancing the development of NLP technologies.

Welcome!

Best,

Aron