CEU Electronic Theses and Dissertations, 2025
| Author | Hgaig, Ghadena |
|---|---|
| Title | Visualizing Political Narratives: A Machine Learning and Knowledge Graph Approach |
| Summary | In this project, I addressed two key challenges in political campaign management: visualizing the media discourse around political actors and improving the efficiency of article classification. I developed an XGBoost classifier using non-LLM features, achieving an 81% recall rate for identifying politically relevant articles. Further gains were observed when focusing on English full-text articles (recall improved to 96%). In parallel, I built an interactive knowledge graph by extracting entities, relationships, and sentiment from the relevant articles. The graph mapped political discourse in the United States, France, and Germany, revealing key insights such as the most central figures in the media landscape and the structure of communities based on media coverage patterns. The knowledge graph enables campaign managers to monitor evolving narratives and analyze connections between key political actors across different countries. While the project has limitations, including language diversity challenges and partial temporal coverage, it demonstrates a viable and scalable solution to equip campaign managers with faster, cheaper, and deeper media insights. |
| Supervisor | Eduardo De La Rubia |
| Department | Economics MSc |
| Full text | https://www.etd.ceu.edu/2025/hgaig_ghadena.pdf |
Visit the CEU Library.
© 2007-2025, Central European University