Scholarlink

Public sentiment toward the U.S. Department of Education (DoE) fluctuates amid ongoing policy changes, student loan debates, and broader trust in federal institutions. While previous work has examined education-related discourse, few studies have leveraged large-scale, cross-platform social media data to assess thematic drivers of trust and engagement. This paper employs natural language processing techniques, including VADER and transformer-based sentiment analysis, alongside Latent Dirichlet Allocation (LDA) and BERTopic modeling, to analyze Reddit and Instagram content from April 2025. Regression and correlation analyses are used to identify patterns linking emotional tone and topical discourse to user interactions across platforms. Results indicate Reddit hosts more policy-oriented and critical discourse, while Instagram emphasizes personal storytelling and advocacy. Topic modeling confirms Reddit’s focus on systemic frustration and Instagram’s storytelling nature. Engagement varies by sentiment and topic, with notable spikes following federal announcements and FAFSA deadlines. Although sentiment polarity was predominantly neutral across platforms, engagement was not significantly predicted by sentiment polarity. These findings underscore the importance of platform context and discourse modality in evaluating public reactions to education policy. We discuss methodological limitations and propose directions for future research that include model refinement and multi-modal sentiment analysis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Public Sentiment Analysis Toward the Department of Education: A Social Media Study Using Topic Modeling and Sentiment Analysis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Irma de la Pena,
  • Manon Pilaud,
  • Ian McCulloh

摘要

Public sentiment toward the U.S. Department of Education (DoE) fluctuates amid ongoing policy changes, student loan debates, and broader trust in federal institutions. While previous work has examined education-related discourse, few studies have leveraged large-scale, cross-platform social media data to assess thematic drivers of trust and engagement. This paper employs natural language processing techniques, including VADER and transformer-based sentiment analysis, alongside Latent Dirichlet Allocation (LDA) and BERTopic modeling, to analyze Reddit and Instagram content from April 2025. Regression and correlation analyses are used to identify patterns linking emotional tone and topical discourse to user interactions across platforms. Results indicate Reddit hosts more policy-oriented and critical discourse, while Instagram emphasizes personal storytelling and advocacy. Topic modeling confirms Reddit’s focus on systemic frustration and Instagram’s storytelling nature. Engagement varies by sentiment and topic, with notable spikes following federal announcements and FAFSA deadlines. Although sentiment polarity was predominantly neutral across platforms, engagement was not significantly predicted by sentiment polarity. These findings underscore the importance of platform context and discourse modality in evaluating public reactions to education policy. We discuss methodological limitations and propose directions for future research that include model refinement and multi-modal sentiment analysis.