Can large language models be used to extract a semantic network physics knowledge base from a specific resource such as a paper?