GBC: GameScript Generation Using BERT Classification: —Enhancing Interactive Narratives with LLM

Generative Artificial Intelligence (GAI) and Large Language Models (LLMs) have been applied in text-based games, but its potential in generating executable game scripts is unexplored. By establishing a mapping from text responses to game scripts, this research explored the application of LLMs in practical game production. This study utilizes a fine-tuned Bidrectional Encoder Representations from Transformers (BERT) model to transform dialogue generated by LLMs into practical game scripts. The proposed method achieves state-of-the-art performance of transformation with an F1 score of 89.71% and an accuracy of 89.74% on the validation set, while maintaining low training and computational costs. The results highlight the potential of LLMs in game development, paving the way for immersive and personalized gaming experiences. Future research could delve deeper into complex and diverse datasets, exploring the integration of multiple labels and numerical values to enhance game interaction and immersion experiences. This research may lay the groundwork for integrating generative AI into game engines in the future, simplifying the production process of adaptive game content.