{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "2f8f9886",
   "metadata": {},
   "source": [
    "# US Attention Data -- Exploration Notebook\n",
    "\n",
    "**Wikipedia pageviews, Google Trends, and GDELT event data tracking global attention in 2025**\n",
    "\n",
    "Author: [Luke Steuber](https://lukesteuber.com) | Bluesky: [@lukesteuber.com](https://bsky.app/profile/lukesteuber.com)\n",
    "\n",
    "Dataset: [lukeslp/us-attention-data](https://huggingface.co/datasets/lukeslp/us-attention-data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "45de0ae5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "import numpy as np\n",
    "from datetime import datetime\n",
    "\n",
    "plt.style.use('seaborn-v0_8-whitegrid')\n",
    "plt.rcParams['figure.figsize'] = (12, 6)\n",
    "plt.rcParams['font.size'] = 11\n",
    "\n",
    "# Load all datasets\n",
    "with open('trends_data.json') as f:\n",
    "    trends = json.load(f)\n",
    "with open('wikipedia_pageviews.json') as f:\n",
    "    wiki_pageviews = json.load(f)\n",
    "with open('gdelt_weekly_events.json') as f:\n",
    "    gdelt_weekly = json.load(f)\n",
    "with open('unified_data.json') as f:\n",
    "    unified = json.load(f)\n",
    "with open('weekly_trends.json') as f:\n",
    "    weekly_trends = json.load(f)\n",
    "with open('wikipedia_trending.json') as f:\n",
    "    wiki_trending = json.load(f)\n",
    "with open('events_unified.json') as f:\n",
    "    events = json.load(f)\n",
    "with open('weekly_attention_timeline.json') as f:\n",
    "    attention_timeline = json.load(f)\n",
    "\n",
    "print(\"Loaded datasets:\")\n",
    "print(f\"  Trends terms:          {len(trends.get('terms', {}))}\")\n",
    "print(f\"  Wikipedia countries:   {len(wiki_pageviews.get('countries', {}))}\")\n",
    "print(f\"  GDELT weekly events:   {len(gdelt_weekly.get('weekly_events', []))}\")\n",
    "print(f\"  Unified countries:     {len(unified.get('countries', []))}\")\n",
    "print(f\"  Weekly trends:         {len(weekly_trends.get('weeks', []))}\")\n",
    "print(f\"  Wikipedia trending:    {len(wiki_trending)}\")\n",
    "print(f\"  Unified events:        {len(events.get('events', []))}\")\n",
    "print(f\"  Attention timeline:    {len(attention_timeline.get('weekly_timeline', []))}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4c9f2e6",
   "metadata": {},
   "source": [
    "## Most Viewed Wikipedia Articles\n",
    "\n",
    "Which articles drew the most attention across all tracked countries?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4b6ee5b7",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_trending = pd.DataFrame(wiki_trending)\n",
    "top_20 = df_trending.nlargest(20, 'total_views')\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(14, 8))\n",
    "bars = ax.barh(range(len(top_20)), top_20['total_views'].values, color='#1565C0')\n",
    "ax.set_yticks(range(len(top_20)))\n",
    "ax.set_yticklabels(top_20['title'].values)\n",
    "ax.set_xlabel('Total Pageviews')\n",
    "ax.set_title('Top 20 Most Viewed Wikipedia Articles (2025)')\n",
    "ax.invert_yaxis()\n",
    "ax.xaxis.set_major_formatter(ticker.FuncFormatter(lambda x, _: f'{x/1e6:.1f}M'))\n",
    "\n",
    "for bar, val in zip(bars, top_20['total_views'].values):\n",
    "    ax.text(bar.get_width() + 1e5, bar.get_y() + bar.get_height()/2,\n",
    "            f'{val/1e6:.1f}M', va='center', fontsize=9)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "print(f\"Total trending articles tracked: {len(wiki_trending)}\")\n",
    "print(f\"Highest daily peak: {df_trending['peak_views'].max():,} views\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "48ce9632",
   "metadata": {},
   "source": [
    "## Google Trends -- Search Interest Over Time\n",
    "\n",
    "What were people searching for throughout 2025?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "15d70f10",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Weekly trends analysis\n",
    "weeks_data = weekly_trends.get('weeks', [])\n",
    "df_weeks = pd.DataFrame(weeks_data)\n",
    "\n",
    "fig, axes = plt.subplots(2, 1, figsize=(16, 10))\n",
    "\n",
    "# Global intensity over time\n",
    "df_weeks['start_date_parsed'] = pd.to_datetime(df_weeks['start_date'])\n",
    "axes[0].plot(df_weeks['start_date_parsed'], pd.to_numeric(df_weeks['global_intensity'], errors='coerce'),\n",
    "             color='#E91E63', linewidth=2, marker='o', markersize=4)\n",
    "axes[0].set_ylabel('Global Intensity')\n",
    "axes[0].set_title('Google Trends -- Global Search Intensity by Week (2025)')\n",
    "axes[0].tick_params(axis='x', rotation=45)\n",
    "\n",
    "# US intensity\n",
    "axes[1].plot(df_weeks['start_date_parsed'], pd.to_numeric(df_weeks['us_intensity'], errors='coerce'),\n",
    "             color='#2196F3', linewidth=2, marker='s', markersize=4)\n",
    "axes[1].set_ylabel('US Intensity')\n",
    "axes[1].set_title('Google Trends -- US Search Intensity by Week')\n",
    "axes[1].set_xlabel('Week')\n",
    "axes[1].tick_params(axis='x', rotation=45)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "# Show top US searches\n",
    "print(\"\\nTop US searches by week (sample):\")\n",
    "for week in df_weeks.head(10).itertuples():\n",
    "    print(f\"  Week {week.week}: {week.top_search_us}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2124bca9",
   "metadata": {},
   "source": [
    "## GDELT Events Timeline\n",
    "\n",
    "The GDELT Project monitors world events from news sources globally."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fc1583b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# GDELT weekly event volume\n",
    "gdelt_weeks = gdelt_weekly.get('weekly_events', [])\n",
    "\n",
    "week_dates = []\n",
    "event_counts = []\n",
    "for w in gdelt_weeks:\n",
    "    week_dates.append(w['week_start'])\n",
    "    event_counts.append(len(w.get('events', [])))\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(16, 6))\n",
    "dates = pd.to_datetime(week_dates)\n",
    "ax.bar(dates, event_counts, width=5, color='#FF5722', alpha=0.8)\n",
    "ax.set_xlabel('Week')\n",
    "ax.set_ylabel('Number of Events')\n",
    "ax.set_title('GDELT -- Weekly Event Volume (2025)')\n",
    "ax.tick_params(axis='x', rotation=45)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "total_events = sum(event_counts)\n",
    "print(f\"Total GDELT events tracked: {total_events:,}\")\n",
    "print(f\"Average events per week: {total_events / len(gdelt_weeks):.1f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "799c4d6e",
   "metadata": {},
   "source": [
    "## Unified Events -- Impact Analysis\n",
    "\n",
    "Major events tracked across all data sources, scored by impact."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2c28fb5f",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_events = pd.DataFrame(events.get('events', []))\n",
    "\n",
    "# Impact score distribution\n",
    "fig, axes = plt.subplots(1, 2, figsize=(16, 7))\n",
    "\n",
    "axes[0].hist(pd.to_numeric(df_events['impact_score'], errors='coerce').dropna(),\n",
    "             bins=20, color='#9C27B0', edgecolor='white', alpha=0.8)\n",
    "axes[0].set_xlabel('Impact Score')\n",
    "axes[0].set_ylabel('Number of Events')\n",
    "axes[0].set_title('Distribution of Event Impact Scores')\n",
    "\n",
    "# Sentiment direction\n",
    "sentiment = df_events['sentiment_direction'].value_counts()\n",
    "colors_sent = {'negative': '#F44336', 'positive': '#4CAF50', 'neutral': '#9E9E9E', 'mixed': '#FF9800'}\n",
    "bars = axes[1].bar(sentiment.index, sentiment.values,\n",
    "                   color=[colors_sent.get(s, '#607D8B') for s in sentiment.index])\n",
    "axes[1].set_ylabel('Number of Events')\n",
    "axes[1].set_title('Events by Sentiment Direction')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "# Top impact events\n",
    "top_events = df_events.nlargest(10, 'impact_score')[['name', 'date', 'impact_score', 'sentiment_direction']]\n",
    "print(\"\\nTop 10 Highest Impact Events:\")\n",
    "for _, row in top_events.iterrows():\n",
    "    print(f\"  [{row['impact_score']}] {row['date']}: {row['name']} ({row['sentiment_direction']})\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0736421c",
   "metadata": {},
   "source": [
    "## Country-Level Attention Comparison\n",
    "\n",
    "How does media/search attention vary across countries in the unified dataset?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ff0fc737",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Unified country data\n",
    "country_list = unified.get('countries', [])\n",
    "df_unified_countries = pd.DataFrame(country_list)\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(14, 7))\n",
    "country_names = df_unified_countries['name'].values\n",
    "regions = df_unified_countries['region'].values\n",
    "\n",
    "region_colors = {}\n",
    "unique_regions = list(set(regions))\n",
    "cmap = plt.cm.Set2(np.linspace(0, 1, len(unique_regions)))\n",
    "for i, r in enumerate(unique_regions):\n",
    "    region_colors[r] = cmap[i]\n",
    "\n",
    "bar_colors = [region_colors[r] for r in regions]\n",
    "bars = ax.barh(range(len(country_names)), range(len(country_names), 0, -1),\n",
    "               color=bar_colors)\n",
    "\n",
    "# Instead, show sentiment timelines for a few key countries\n",
    "fig2, ax2 = plt.subplots(figsize=(16, 7))\n",
    "highlight_countries = ['US', 'GB', 'FR', 'DE', 'CA', 'MX']\n",
    "for c_data in country_list:\n",
    "    if c_data['code'] in highlight_countries:\n",
    "        timeline = c_data.get('sentiment_timeline', [])\n",
    "        if timeline:\n",
    "            dates = [t.get('date', t.get('week_start', '')) for t in timeline]\n",
    "            scores = [t.get('score', t.get('sentiment', 0)) for t in timeline]\n",
    "            if dates and scores:\n",
    "                ax2.plot(pd.to_datetime(dates[:52]), [float(s) if s else 0 for s in scores[:52]],\n",
    "                         label=c_data['name'], linewidth=2, marker='o', markersize=3)\n",
    "\n",
    "ax2.set_xlabel('Date')\n",
    "ax2.set_ylabel('Sentiment Score')\n",
    "ax2.set_title('Sentiment Toward US -- Key Countries Over Time')\n",
    "ax2.legend(loc='best')\n",
    "ax2.tick_params(axis='x', rotation=45)\n",
    "ax2.axhline(y=0, color='gray', linestyle='--', alpha=0.5)\n",
    "plt.tight_layout()\n",
    "plt.close(fig)  # close the placeholder\n",
    "plt.show()\n",
    "\n",
    "print(f\"Countries tracked: {len(country_list)}\")\n",
    "print(f\"Regions: {', '.join(unique_regions)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1243e131",
   "metadata": {},
   "source": [
    "## Weekly Attention Timeline\n",
    "\n",
    "Composite attention signal combining Wikipedia, Trends, and GDELT."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e4926681",
   "metadata": {},
   "outputs": [],
   "source": [
    "timeline = attention_timeline.get('weekly_timeline', [])\n",
    "\n",
    "weeks_parsed = []\n",
    "for w in timeline:\n",
    "    entry = {'week_start': w['week_start']}\n",
    "    components = w.get('components', {})\n",
    "    for key in components:\n",
    "        if isinstance(components[key], (int, float)):\n",
    "            entry[key] = components[key]\n",
    "        elif isinstance(components[key], dict):\n",
    "            for subkey, val in components[key].items():\n",
    "                if isinstance(val, (int, float)):\n",
    "                    entry[f'{key}_{subkey}'] = val\n",
    "    weeks_parsed.append(entry)\n",
    "\n",
    "df_timeline = pd.DataFrame(weeks_parsed)\n",
    "df_timeline['week_start'] = pd.to_datetime(df_timeline['week_start'])\n",
    "\n",
    "# Plot available numeric columns\n",
    "numeric_cols = [c for c in df_timeline.columns if c != 'week_start' and df_timeline[c].dtype in ['float64', 'int64']]\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(16, 7))\n",
    "for i, col in enumerate(numeric_cols[:5]):\n",
    "    ax.plot(df_timeline['week_start'], df_timeline[col],\n",
    "            label=col.replace('_', ' ').title(), linewidth=1.5, alpha=0.8)\n",
    "\n",
    "ax.set_xlabel('Week')\n",
    "ax.set_ylabel('Score')\n",
    "ax.set_title('Weekly Attention Components Over Time')\n",
    "ax.legend(loc='best', fontsize=9)\n",
    "ax.tick_params(axis='x', rotation=45)\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "print(f\"Weeks tracked: {len(timeline)}\")\n",
    "print(f\"Components available: {numeric_cols[:10]}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e3d26c53",
   "metadata": {},
   "source": [
    "## Summary Statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "60073285",
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"=\" * 60)\n",
    "print(\"US ATTENTION DATA -- SUMMARY\")\n",
    "print(\"=\" * 60)\n",
    "print(f\"Timeframe:               {trends.get('timeframe', 'N/A')}\")\n",
    "print(f\"Wikipedia trending:      {len(wiki_trending):>10,} articles\")\n",
    "print(f\"GDELT weeks tracked:     {len(gdelt_weekly.get('weekly_events',[])):>10}\")\n",
    "print(f\"Unified events:          {len(events.get('events',[])):>10}\")\n",
    "print(f\"Countries tracked:       {len(unified.get('countries',[])):>10}\")\n",
    "print(f\"Weekly trends:           {len(weekly_trends.get('weeks',[])):>10}\")\n",
    "print(f\"Search terms tracked:    {len(trends.get('terms',{})):>10}\")\n",
    "print()\n",
    "print(\"Data sources:\")\n",
    "for src in unified.get('metadata', {}).get('data_sources', []):\n",
    "    if isinstance(src, dict):\n",
    "        print(f\"  -- {src.get('name', src)}: {src.get('description', '')}\")\n",
    "    else:\n",
    "        print(f\"  -- {src}\")\n",
    "print(\"=\" * 60)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}