Are Large Language Models Effective Knowledge Graph Constructors?