Abstract

Large language models are powerful tools for story generation, but are difficult to control. Story grammars are a more controllable tool for story generation, but require a large amount of upfront work and tend to create predictable results. We present a story generation tool that combines the positives of both methods using language models to aid in writing a grammar, and using the output of that grammar to generate controlled text from language models. Our approach combines the structure of a grammar with the unexpectedness of a large language model.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call