Nested entities in named entity recognition (NER) present a significant challenge in natural language processing. This paper proposes a new annotation rule for sequence tagging models, developed based on the identified patterns within nested entities. We applied this annotation method to other models, such as using it as a prompt for ChatGPT, to demonstrate the rule's universality. This study not only optimizes the annotation process for sequence tagging models but also validates the universality of the method in improving nested entity recognition.