News Analytics
FastAPI-backed source, lens, workflow, and event diagnostics
The news surface reads from the backend contract and exposes exploratory comparisons, data quality checks, workflow health, and same-event analysis.
Index
Pages
All routes below stay inside the same shared analytics surface.
News Lens PCA
Live FastAPI data
What this page does
View latent lens components that explain the largest variance in the dataset.
Data Mode
Snapshot mode requires a valid date (`YYYY-MM-DD`). Falling back to current data until provided.
Overview
Latent Space Status
Coverage and fit diagnostics for the PCA, MDS, and source-separation views.
ok
Charts
PCA Visuals
Variance allocation plus article-level geometry in both PCA and distance-preserving MDS space.
Reference Table
Explained Variance
Eigenvalue contribution and cumulative variance captured by each component.
| Component | Eigenvalue | Explained | Cumulative |
|---|---|---|---|
| PC1 | 3.4679 | 28.7% | 28.7% |
| PC2 | 1.6892 | 14.0% | 42.7% |
| PC3 | 1.2824 | 10.6% | 53.4% |
| PC4 | 0.9666 | 8.0% | 61.4% |
| PC5 | 0.8928 | 7.4% | 68.8% |
| PC6 | 0.7733 | 6.4% | 75.2% |
Loadings
Variance Drivers
The lenses contributing most strongly to the first two principal components.
| Lens | Weighted Contribution | PC1 | PC2 |
|---|---|---|---|
| Linguistic Quality Lens | 0.0780 | -0.0615 | 0.0746 |
| Objectivity vs Opinion Lens | 0.0707 | 0.1454 | -0.4302 |
| Entity-Level Sentiment Lens | 0.0683 | -0.0481 | 0.4951 |
| Agency and Voice Lens | 0.0645 | 0.2795 | 0.1279 |
| Omission and Silence Lens | 0.0644 | 0.3410 | 0.0771 |
| Hegemonic Common-Sense Lens | 0.0641 | 0.3272 | -0.1772 |
| Emotional Intensity Lens | 0.0627 | 0.2967 | 0.4032 |
| Sentiment Clarity Lens | 0.0615 | 0.0755 | 0.5537 |
| Epistemic Modality and Certainty Lens | 0.0571 | 0.3272 | -0.1041 |
| Authority and Source Positioning Lens | 0.0553 | 0.3371 | -0.1693 |
| Credibility Lens | 0.0526 | 0.4206 | -0.0039 |
| Causal Attribution Lens | 0.0524 | 0.4243 | 0.0428 |
Source Geometry
Source Centroids
Average source positions inside the leading component space.
| Source | Count | PC1 | PC2 |
|---|---|---|---|
| BBC News | 61 | -0.034 | -0.243 |
| NPR | 51 | -0.332 | 0.090 |
| The Guardian | 40 | 1.220 | 0.190 |
| Fox News | 11 | -1.949 | 0.830 |
| PBS NewsHour | 5 | -1.311 | -0.567 |
| CBS News | 2 | 0.318 | -1.429 |
| ABC News | 1 | -0.184 | -1.219 |
| NBC News | 1 | -2.265 | 0.402 |
Diagnostics
Source Separation Diagnostics
Silhouette-like source separation measures for the latent representation.
ok
| Source | Articles | Silhouette-Like Mean |
|---|---|---|
| BBC News | 61 | -0.272 |
| NPR | 51 | -0.243 |
| The Guardian | 40 | -0.075 |
| Fox News | 11 | -0.241 |
| PBS NewsHour | 5 | -0.167 |
| CBS News | 2 | -0.388 |