Transfer learning and in-context learning for stage direction classification in French

This poster expands our work on stage direction classification in French via fine-tuning pre-trained language models and prompting large language models (LLM), testing new models, hyperparameters and prompts. A new qualitative analysis of LLM results showed limits in our reference annotations, and how LLMs can help identify them.

Palabras clave: Transfer Learning, Prompting, Large Language Models, Computational Drama Analysis,