Generating short RPG side-quest stories with transformers

Progress in natural language processing has allowed for the exploration of automatically creating creatively written stories. The proposed research will exploit that possibility. By using GPT-2 trained additionally on a purpose build data-set the system will generate short stories to support RPG side-quests. This paper’s contribution is to show that fine-tuning a general pretrained transformer for story generation has potential to be a good option for obtaining descriptions for quests, if used in computer-human collaboration. It also highlights the shortcomings in the current method and proposes how those shortcomings may be addressed in future attempts.