LLM Based Public Message Refinedly Grading Method

In order to grade the urgency of public messages, this paper presents a LLM Based Public Message Refinedly Grading Method. This method designs a set of key factors for message classification, including sentiment polarity and type, as well as event type and domain. It utilizes pre-trained large-scale language models to identify these factors in the message content. Through a genetic algorithm, the weights of various factors are optimized for comprehensive scoring, enabling automated message classification. Experimental results demonstrate that this approach effectively improves the accuracy of message classification.