Public Message Analysis Based on LDA-BERT-BiLSTM -- Taking the “Leadership Message Board” of People’s Daily Website as an Example

[Objective/Significance] This article explores the theme of elderly care and its emotional trend, providing a scientific basis for formulating and optimizing elderly care policies, thereby improving the quality of elderly care services and promoting the harmonious and stable development of society. [Method/Process] This study uses the message data related to pension issues in the “leadership message board” platform of People’s Daily Online as the research object, the LDA model to extract relevant topics, and the BERT-BiLSTM model to analyze the emotional trend changes of each subject. [Result/Conclusion] The research results show that the theme of pension covers 16 sub-subjects, such as “environmental noise problem of pension,” “pension community construction planning,” and “retirement pension policy treatment.” The classification accuracy of the BERT-BiLSTM model is 0.89, and the F1 value is 0.86, which shows its superiority in sentiment evolution analysis. The expression of negative emotions in pension theme messages is generally higher than that of neutral and positive feelings, and the elderly group’s emotional attitude shows significant differences under different themes of raising older people.