Controversy Detection¶
Controversy detection on social media aims to identify topics, discussions, or events that spark heated or polarized debate. Methods range from content-based approaches (sentiment analysis, language patterns) to network-based techniques that analyze the structure of user interactions and graph partitioning to identify opposing sides.
Key finding¶
Network structure is more reliable than content for controversy detection: graph-based measures that analyze conversation topology—specifically the degree of clustering and separation between user communities—outperform text-only approaches and generalize across diverse topics and platforms.
Key papers¶
- Garimella et al. (2017) — Quantifying Controversy on Social Media — Three-stage pipeline combining graph construction, graph partitioning, and controversy measurement via random-walk-based metrics; demonstrates RWC metric outperforms existing baselines on Twitter and external datasets.
- Garimella et al. (2016) — Reducing Controversy by Connecting Opposing Views — Uses RWC metric to formulate an optimization problem: select k edges to add that minimize expected controversy; proposes efficient algorithm considering only high-degree hub nodes across opposing sides.