QuestVille: Procedural Quest Generation Using NLP Models

Abstract

Developers face a time-consuming task when creating quests in video games. To ease this burden, Procedural Content Generation (PCG) techniques can be used to automatically generate quests. While PCG has been applied to various areas of game development, it can be difficult to create meaningful narratives for quests. This paper presents a new method for generating engaging quests by combining PCG with Natural Language Processing (NLP) using the models BERT and GPT-2 in a case study game called QuestVille. The paper details the implementation of these models and the challenges encountered. The results suggest that the use of BERT and GPT-2 has potential for creating compelling narrative content. Advancements in AI research may improve on the limitations discussed.