The task in this paper is to extract alias relations from general sentences. It can be used to expand the knowledge base and provide support to entity linking. The problem with traditional alias relation extraction methods is that they work better for identifying frequent alias expressions than infrequent ones. In order to solve this problem, we propose to search keywords for alias expressions with the help of the “XIANDAI HANYU CIDIAN” (Contemporary Chinese Dictionary) and try to cover as many infrequent alias expressions as possible. We generate training samples for the infrequent keywords using large language models automatically to enlarge the dataset. The experimental results show that the model performs better when trained based on the enlarged training set with the generated instances.