A Case Grammar-Based Knowledge Structure Extraction Method for Story Generation Using Machine Learning

Abstract

In recent years, the development of general-purpose large-scale language models, including GPT, has been promoted by OpenAI, and in the field of narrative generation, there have been many proposals for models that retrain trained GPT2. However, there is a problem that it is difficult to generate sentences that take into account events prior to a certain time in the generated story. In this paper, we propose a method for extracting knowledge structures based on case grammars from sentences in a dataset by analyzing dependencies between words.