Park Message Recommendation System Based on Large Language Model And Evaluation Metrics

With the continuous development of the park, it is necessary for the park managers to get the information related to the development of the park in time, so it becomes more important to build a message recommendation system for the managers of science and technology parks. We propose an message recommendation algorithm that combines the large language model and the park evaluation metrics, and there are two main innovations: one is to fine-tune the large language model to do the message representation, so as to synthesize the open general knowledge of the large language model and improve the model's representation of the message; the other is to embed the evaluation information of the users into the user's representation, analyze the data of the parks, and derive the labels of each park using the hierarchical analysis method. Then we use the label to generate new message for the user's representation, which makes the representation of the user more accurate. Finally, we conducted experiments on the paper and patent dataset, which showed that using the algorithm can improve the accuracy of the recommendation system.