Little Red Riding Hood Goes Around the Globe:Crosslingual Story Planning and Generation with Large Language Models

TL;DR

This paper explores cross-lingual story generation using planning with large language models, introducing a new dataset and demonstrating that structured plans improve story coherence and controllability across languages.

Contribution

It introduces a novel cross-lingual story planning and generation task, along with a new dataset, and shows that structured three-act plans enhance story quality and control.

Findings

Three-act plans improve story coherence

Planning enables explicit content and structure control

Large language models effectively generate multi-language stories

Abstract

Previous work has demonstrated the effectiveness of planning for story generation exclusively in a monolingual setting focusing primarily on English. We consider whether planning brings advantages to automatic story generation across languages. We propose a new task of cross-lingual story generation with planning and present a new dataset for this task. We conduct a comprehensive study of different plans and generate stories in several languages, by leveraging the creative and reasoning capabilities of large pre-trained language models. Our results demonstrate that plans which structure stories into three acts lead to more coherent and interesting narratives, while allowing to explicitly control their content and structure.