Datasets:

lukeslp
/

us-attention-data

@@ -1,92 +1,164 @@
 # US Attention Data
-[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
-[![Data Sources](https://img.shields.io/badge/Sources-Wikipedia%20%7C%20GDELT%20%7C%20Google%20Trends-green)](https://en.wikipedia.org/wiki/Special:Statistics)
-[![HuggingFace](https://img.shields.io/badge/%F0%9F%A4%97-HuggingFace-yellow)](https://huggingface.co/datasets/lukeslp/us-attention-data)
-Weekly cross-platform attention metrics for tracking how much the world pays attention to the United States. Combines Wikipedia pageviews, GDELT global event mentions, and Google Trends search interest from 2020-2025.
-I built this dataset for the [one-year](https://dr.eamer.dev/datavis/one-year/) visualization project, which maps US global sentiment over time. Part of the [Data Trove](https://dr.eamer.dev/datavis/data_trove/) collection.
----
-## What's Inside
-| File | Size | Description |
-|------|------|-------------|
-| `wikipedia_pageviews.json` | 2.5 MB | Daily pageview counts for US-related Wikipedia articles |
-| `wikipedia_event_articles.json` | 214 KB | Event-linked article metadata |
-| `wikipedia_trending.json` | 256 KB | Trending article detection |
-| `trends_data.json` | 810 KB | Google Trends search interest over time |
-| `weekly_trends.json` | 26 KB | Weekly trending topic aggregations |
-| `gdelt_timeline.json` | 131 KB | GDELT event mention timelines |
-| `gdelt_weekly_events.json` | 158 KB | GDELT weekly aggregated event counts and tone |
-| `events_unified.json` | 89 KB | Unified event data across all sources |
-| `weekly_attention_timeline.json` | 57 KB | Combined weekly attention metrics |
-| `unified_data.json` | 27 KB | Merged dataset across all attention sources |
-| `attention_metadata.json` | 2 KB | Collection metadata and schema |
-**Total: ~4.2 MB**
----
-## Quick Start
-### Python
 ```python
 import json
-with open("wikipedia_pageviews.json") as f:
-    pageviews = json.load(f)
-# Weekly attention across all sources
-with open("weekly_attention_timeline.json") as f:
-    timeline = json.load(f)
-```
-### D3.js
-```javascript
-const pageviews = await d3.json("wikipedia_pageviews.json");
-const gdelt = await d3.json("gdelt_weekly_events.json");
 ```
----
-## Data Sources
-| Source | What It Tracks | Coverage |
-|--------|---------------|----------|
-| [Wikipedia Pageviews API](https://wikimedia.org/api/rest_v1/) | Article view counts | 2020-2025, daily |
-| [GDELT Project](https://www.gdeltproject.org/) | Global event mentions and media tone | 2020-2025, weekly |
-| [Google Trends](https://trends.google.com/) | Search interest indices | 2020-2025, weekly |
----
-## Use Cases
-- Tracking how global attention to the US shifts over time
-- Correlating media events with Wikipedia traffic and search interest
-- Identifying seasonal attention patterns (elections, holidays, crises)
-- Building composite attention indices from multiple independent signals
----
-## Related
-- [one-year visualization](https://dr.eamer.dev/datavis/one-year/) -- the viz this data powers
-- [Data Trove](https://dr.eamer.dev/datavis/data_trove/) -- full dataset catalog
-- [lukesteuber.com](https://lukesteuber.com) -- portfolio
----
-## Author
-[Luke Steuber](https://lukesteuber.com) -- [@lukesteuber.com](https://bsky.app/profile/lukesteuber.com) on Bluesky
-## License
-MIT. See [LICENSE](LICENSE).
-Data sourced from Wikipedia (CC BY-SA), GDELT (open), and Google Trends (fair use for research).

+---
+license: mit
+task_categories:
+- time-series-forecasting
+- text-classification
+language:
+- en
+tags:
+- attention-metrics
+- wikipedia
+- google-trends
+- gdelt
+- pageviews
+- search-interest
+- media-coverage
+- sentiment-analysis
+- event-tracking
+pretty_name: US Attention Data
+size_categories:
+- n<1K
+---
 # US Attention Data
+Digital attention metrics tracking Wikipedia pageviews, Google Trends search interest, and GDELT global event coverage for United States-related topics throughout 2025.
+## Dataset Description
+This dataset aggregates attention signals from three authoritative sources to track how digital attention flows around US-related topics:
+- **Wikipedia Pageviews** (Wikimedia REST API) - 2.5MB tracking 8 key articles across countries
+- **Google Trends** (pytrends API) - 810KB covering 138 search terms in the US region
+- **GDELT 2.1** (Doc API) - Event mentions across 44 countries
+### Use Cases
+- Media attention analysis
+- Event impact measurement
+- Search trend correlation
+- Cross-platform attention patterns
+- News cycle dynamics
+- Public interest tracking
+## Dataset Structure
+### Data Files (10 total, 4.3 MB)
+#### Wikipedia Pageviews
+- `wikipedia_pageviews.json` - 2.5 MB - Country-level pageview analysis
+- `wikipedia_trending.json` - 250 KB - Top 500 trending articles
+- `wikipedia_event_articles.json` - 210 KB - Event-driven article tracking
+#### Google Trends
+- `trends_data.json` - 810 KB - 138 search terms, full year time series
+- `weekly_trends.json` - 26 KB - Weekly aggregated snapshot
+#### GDELT Event Coverage
+- `gdelt_timeline.json` - 131 KB - 44 countries, event mentions
+- `gdelt_weekly_events.json` - 158 KB - Weekly event aggregation
+#### Unified Cross-Source
+- `events_unified.json` - 88 KB - 177 events combined
+- `unified_data.json` - 27 KB - High-level unified analysis
+- `weekly_attention_timeline.json` - 56 KB - Weekly cross-platform metrics
+### Metadata Files
+Each data file has a companion `*_metadata.json` with:
+- Source API details
+- Field descriptions
+- Update timestamps
+- Record counts
+- Usage notes
+## Data Collection
+### Wikipedia (Wikimedia REST API)
+- **Endpoint**: `https://wikimedia.org/api/rest_v1`
+- **Auth**: None required
+- **Rate Limit**: 5000 req/hour per IP
+- **License**: CC0 Public Domain
+### Google Trends (pytrends)
+- **Package**: `pytrends` Python library
+- **Auth**: None required
+- **Rate Limit**: Soft limits (use responsibly)
+- **License**: Google Terms of Service
+### GDELT (GDELT 2.1 Doc API)
+- **Endpoint**: `https://api.gdeltproject.org/api/v2/doc/doc`
+- **Query**: `("United States" OR "United States of America")`
+- **Auth**: None required
+- **License**: Free for research and commercial use
+## Loading the Data
 ```python
 import json
+import pandas as pd
+# Load Wikipedia pageviews
+with open('data/wikipedia_pageviews.json') as f:
+    wp_data = json.load(f)
+# Load Google Trends
+trends = pd.read_json('data/trends_data.json')
+# Load GDELT events
+with open('data/gdelt_timeline.json') as f:
+    gdelt = json.load(f)
 ```
+## Key Findings
+- **Wikipedia** serves as primary information source with event-driven pageview spikes
+- **Google Trends** shows search volatility averaging 3-5 day persistence
+- **GDELT** reveals uneven global media coverage (Western bias)
+- **Cross-source correlation** during major events (disasters, elections, conflicts)
+- **Attention decay** follows exponential curve with 2-4 day half-life
+## Visualizations
+Three live dashboards built on this data:
+1. **One-Year US Global Sentiment** - [dr.eamer.dev/datavis/one-year](https://dr.eamer.dev/datavis/one-year)
+   - Uses: `gdelt_timeline.json`
+   - Visualizes global media coverage patterns
+2. **Trends 2025** - [dr.eamer.dev/datavis/trends-2025](https://dr.eamer.dev/datavis/trends-2025)
+   - Uses: `trends_data.json`
+   - Interactive Google Trends dashboard
+3. **Wikipedia Attention Analytics** - [dr.eamer.dev/datavis/wiki-attention](https://dr.eamer.dev/datavis/wiki-attention)
+   - Uses: `wikipedia_pageviews.json`
+   - Pageview trend analysis
+## Citation
+```bibtex
+@dataset{steuber2026attention,
+  title={US Attention Data: Digital Attention Metrics from Wikipedia, Google Trends, and GDELT},
+  author={Steuber, Luke},
+  year={2026},
+  publisher={HuggingFace},
+  url={https://huggingface.co/datasets/lukeslp/us-attention-data}
+}
+```
+## License
+MIT License - See LICENSE file for details.
+Data sources have their own licenses:
+- Wikipedia data: CC0 Public Domain
+- Google Trends: Subject to Google Terms of Service
+- GDELT: Free for research and commercial use
+## About
+Created by Luke Steuber
+- **Website**: [lukesteuber.com](https://lukesteuber.com)
+- **Portfolio**: [dr.eamer.dev](https://dr.eamer.dev)
+- **Bluesky**: [@lukesteuber.com](https://bsky.app/profile/lukesteuber.com)