Text this: Enhancing large language models for scientific entity recognition via fully automated self-distillation