Back to top

Conclusion

  • Introduction
  • Exploratory data analysis
  • Data Visualization
  • Methodology
  • Sentiment Analysis
  • Conclusion
  • Code

    • Conclusion

    The analysis demonstrates that rumor dissemination exhibits polycentric diffusion patterns, primarily driven by multiple influential user accounts acting as independent propagation hubs. These key nodes facilitate content amplification across distinct subnetworks, enabling parallelized spread. Concurrently, the propagation follows a hierarchical structure, where information cascades from primary disseminators to secondary recipients before reaching peripheral audiences. This dual mechanism—combining multi-center initiation with tiered redistribution—explains the accelerated velocity and expanded reach characteristic of rumor diffusion compared to fact-based information flows.

    Future Study

    To enhance the analytical framework, four priorities are proposed:

    1. Domain-Adapted Language Models: Training a dedicated BERT model on annotated platform-specific corpora to improve the semantic capture of colloquial and context-dependent expressions.
    2. Sentiment Analysis Optimization: Investigating architectures beyond baseline transformers for better alignment with nuanced emotional lexicons in user-generated content.
    3. Proactive Rumor Detection: Developing classifiers trained on discriminative textual features (e.g., speculative phrasing, urgency markers) to predict rumor proliferation risks.
    4. Affective Trigger Mapping: Sustained monitoring of emerging topics correlated with negative sentiment spikes to enable preemptive public communication strategies.