A Comprehensive Evaluation of Semantic Relation Knowledge of Pretrained Language Models and Humans