Collect articles
Gather reporting from multiple news organizations.
Computational media analysis
Understand the narratives behind the headlines.
MediaScape compares how different news organizations frame the same issue using discourse analysis, NLP, and data visualization.
Framing overview
Climate Change
Prototype comparison across outlets.
Volume Over Time
Articles published per day, demonstration data
Demonstration dataset
A static prototype with transparent mock data.
How MediaScape works
From raw articles to reliable insights in four steps.
Analyze framing
Detect linguistic frames, rhetorical choices, tone, and recurring themes.
Compare narratives
Examine how outlets emphasize different dimensions of the same issue.
Visualize patterns
Reveal trends, differences, and clusters through interactive visualizations.
Outlet comparison
See how outlets frame the same story.
Compare key framing dimensions side by side across leading news organizations.
- Quantitative framing scores
- Normalized for comparison
- Designed for transparent interpretation
Climate Change
Comparison across outlets
Language and framing analysis
What language shapes the narrative?
Explore the words and rhetorical frames that outlets use most frequently when discussing the same topic.
Framing Types
Share of articles using each dominant frame
Frames are estimated in this prototype using a research-informed classification framework. Production results will require validation against human-coded samples.
Narrative map
Explore the landscape of narratives.
Articles with similar framing and language appear closer together, helping reveal clusters and differences across outlets.
- Each point represents an article
- Closer points suggest similar narratives
- Clusters show dominant storylines
Narrative Map (Semantic Embeddings)
30-day window, demonstration points
Our methodology
Grounded in media studies, linguistics, and climate communication research.
NLP
Identifies language patterns, entities, and lexical choices.
Framing Analysis
Applies a theory-driven classification framework.
Sentiment
Estimates tone and emotional intensity.
Embeddings
Maps semantic similarity across articles.
Topic Modeling
Surfaces recurring themes at scale.
MediaScape is designed as an interpretive research tool, not an automatic authority on political bias or factual truth. Model outputs should be documented, validated, and presented with uncertainty.
Explore the media landscape more clearly.
Start with the climate-change demonstration and see how narratives shape public understanding.