Framing detection¶
Detection and analysis of media frames—the conceptual lenses through which news stories are presented. Framing shapes public perception and can be used strategically to mislead audiences about the nature, causes, and solutions to issues.
Approaches¶
Taxonomy-based detection:
Classifying news articles against established media frame taxonomies (e.g., economic, morality, health impact frames) using linguistic patterns and NLP methods.
Multi-label classification:
Recognizing that single articles often employ multiple frames simultaneously; using multi-label architectures to capture frame complexity.
Explainability:
Generating explanations for detected frames to help researchers understand how language instantiates particular framings.
Cross-domain framing:
Analyzing how the same event is framed differently across political perspectives, outlets, or time periods.
Key papers in this wiki¶
- DELL: Generating Reactions and Explanations for LLM-Based Misinformation Detection (2024) — Includes framing detection as one of three misinformation tasks; uses LLM-generated explanations and expert ensemble to achieve improvements on SemEval-23P and MFC datasets.
Related concepts¶
- Propaganda and coordinated campaigns — related persuasion techniques
- Misinformation and fake news detection — broader detection methods
- Media literacy and critical evaluation — how audiences understand framing