Topics generated by LDA analysis.

The “Double reduction” policy in China was promulgated and implemented on July 24, 2021. This was the most rigorous and radical education reform since China was founded, aimed at easing pupils’ academic burden and improving their physical and mental health. It sparked extensive public debate across stakeholders (schools, teachers, parents). To delve into public sentiment and improve policy management, this paper analyzes Sina Weibo (Chinese Twitter) posts from July to November 2021. We apply text mining to collect microblogs and comments, and we use Latent Dirichlet Allocation (LDA) for topic extraction. A natural language processing-based (NLP) approach is used to analyze the sentiment. Specifically, we fine-tune a pre-trained Chinese BERT model to classify sentiment, leveraging its superior performance on Weibo data. We report sentiment polarity (positive/negative/neutral) and track its trends over time. We also incorporate engagement metadata (likes/reposts) as indicators. Key findings include the distribution of sentiments, major public concerns, and how sentiment levels evolved around key dates. These insights can help policymakers understand public reactions and guide more responsive education reforms.