Modeling Political Discourse with Sentence-BERT and BERTopic

TL;DR

This work tackles how political discourse on Twitter evolves over the 117th U.S. Congress by integrating BERTopic topic modeling with Moral Foundations Theory to quantify topic longevity and moral framing. It introduces a monthly topic-evolution framework and uses a SBERT-based BERTopic pipeline to extract topics, paired with cosine-similarity-based longevity measures and MF scoring via moralstrength. The study finds that broad themes persist while granular topics are short-lived, and that topics with stronger Care and Loyalty framing tend to endure longer, with clear partisan differences in moral framing. The approach offers a scalable, interpretable way to analyze moral-driven topic evolution on social media, with potential for cross-congress comparisons and enhanced engagement-aware insights.

Abstract

Social media has reshaped political discourse, offering politicians a platform for direct engagement while reinforcing polarization and ideological divides. This study introduces a novel topic evolution framework that integrates BERTopic-based topic modeling with Moral Foundations Theory (MFT) to analyze the longevity and moral dimensions of political topics in Twitter activity during the 117th U.S. Congress. We propose a methodology for tracking dynamic topic shifts over time and measuring their association with moral values and quantifying topic persistence. Our findings reveal that while overarching themes remain stable, granular topics tend to dissolve rapidly, limiting their long-term influence. Moreover, moral foundations play a critical role in topic longevity, with Care and Loyalty dominating durable topics, while partisan differences manifest in distinct moral framing strategies. This work contributes to the field of social network analysis and computational political discourse by offering a scalable, interpretable approach to understanding moral-driven topic evolution on social media.

Figures (8)

• Figure 1: BERTopic algorithm
• Figure 2: Examples of topic evolution patterns: (a) short-lived topic, (b) topic with intermittent activity, (c) topic splitting into two, and (d) long-lived merged topic.
• Figure 3: Timeline of topic evolution across the 24-month period. Each horizontal bar represents a topic, spanning the months during which the topic was active. Long bars indicate persistent topics (broad issues recurring over time), whereas short bars represent ephemeral topics tied to transient events.
• Figure 4: Distribution of moral foundation scores for Democratic and Republican legislators. Violin plots show distinct partisan emphasis: Democrats exhibit higher scores on Care and Fairness, while Republicans show elevated scores on Loyalty, Authority, and Purity.
• Figure 5: Distribution of moral foundation scores across topics grouped by longevity. Long-lived topics tend to have higher Care and Loyalty scores, while short-lived topics exhibit weaker moral framing.