The Detection and Understanding of Fictional Discourse

In this paper, we present a variety of classification experiments related to the task of fictional discourse detection. We utilize a diverse array of datasets, including contemporary professionally published fiction, historical fiction from the Hathi Trust, fanfiction, stories from Reddit, folk tales, GPT-generated stories, and anglophone world literature. Additionally, we introduce a new feature set of word "supersenses" that facilitate the goal of semantic generalization. The detection of fictional discourse can help enrich our knowledge of large cultural heritage archives …

array classification cs.cl cs.lg datasets detection discourse diverse fanfiction fiction generated gpt literature paper reddit stories trust understanding world