Exploring Prompt-Based Methods for Zero-Shot Hypernym Prediction with Large Language Models