Fine-Tuned SetFit Models for Classifying Sentences from French Novels into Character and Narrator Speech

Fine-Tuned SetFit Models for Classifying Sentences from French Novels into Character and Narrator Speech. One optimized for first-person (autodiegetic, to some extent homodiegetic) narrative, one optimized for third-person (heterodiegetic, to some extent homodiegetic) narrative. Performance is around 89% accuracy for the first-person model and 93% for the third-person model.