Today we’re tremendously excited to unveil a long-awaited visualization that explores how the world’s news media groups countries into distinct clusters, creating an inherent geographic network structure over the planet akin to “communities” as seen through the eyes of the world’s presses. In essence, for every monitored news article published anywhere in the world that mentions a given country, we compile a list of all other countries also mentioned in those articles. The final result is a ranked list (a co-occurrence histogram) for each country that lists, in order, every other country in the world and how often it is mentioned in coverage about the given country. It is important to note that this is not how often coverage FROM a country mentions other countries, it is how often coverage from any country in the world that is ABOUT that country mentions other countries – in other words, its about context. Countries which are frequently mentioned together might reflect geographic proximity (neighboring countries are more likely to be mentioned together), but also economic and political ties, joint involvement in a major international event or narrative, or contextualization. Examining ties over time can yield a powerful chronology of world events. Ties that last only for a short time period can reflect sudden events such as a natural disaster that prompts coverage in each country of domestic ties to the stricken country, interstate tension and conflict, or even a competitions.
The result is a rich interactive visualization combining GDELT’s data with the Google Arms Globe Visualization to provide a glimpse of how the countries of the world are grouped together in the global press, by week (Sunday to Saturday) from the week of February 22-28, 2015 through the week of May 24-30, 2015. Use your mouse to click and drag the globe to spin it around and use your mouse scrollwheel or the +/- zoom buttons at top-right of the visualization to zoom in/out of the globe. Click on any country (or type its name into the country search box at the top-right of the visualization to select it) to see the countries it co-occurs with. Blue incoming lines indicate countries where at least 10% of coverage mentioning the source country also mentioned the selected country. Red outgoing lines indicate countries where at least 10% of the coverage mentioning the selected country also mentioned the destination country. The size and number of energy pulses moving along the line indicate the percentage of co-occurring mentions (the strength of the connection between them as seen through the world’s collective news media). In a given week nearly every country is mentioned alongside of every other country at least once, so we use the cutoff of 10% to highlight only the strongest co-occurrence relationships in the visualization.
Once you select a country, all countries connected to it will be highlighted in light grey. Zooming into the globe over one of those countries will display a popup with its name – hovering your mouse over that popup will display the percent of the selected country’s mentions that mentioned this country (“cooccurrences in”) and the percent of this country’s mentions that also mentioned the selected country (“cooccurrences out”). By using percentages instead of raw article counts the visualization is able to normalize for the fact that there are simply more mentions globally of the United States than there are of Suriname, for example. Note that for the selected country this means that the percentages it displays are a sum total of all percentages of countries that co-occur with it, meaning you will see numbers in excess of 100%. You can turn incoming/outgoing cooccurrences on/off using the checkboxes at the bottom right of the visualization.
Look carefully at the connections among countries and you’ll see a network chronology of world history over the last three months week-by-week. Watching how a country’s ties change over time can yield fascinating results. Select Yemen and move the time slider from left to right, noting how the country suddenly attracts considerable international attention the week of March 22nd, with the launch of the Saudi air strikes against the Houthi rebels there. If you turn off outgoing links via the checkbox at the bottom right of the visualization, you can see especially clearly the massive jump in the number of blue incoming links that week. Similarly, select the Falkland Islands and click on the graph icon at the bottom of the visualization, just to the right of the week timeline slider. Notice the sharp increase in countries co-occurring with the Falkland Islands the week of March 29th through the week of April 26th as tensions surged between the UK and Argentina over the islands after the announcement on April 2nd of significant oil and gas reserves discovered off their coast. Select Nepal and you can see the intense global focus on the nation after its earthquake. Select Estonia and you will see a curious tie with Argentina the week of February 22 that disappears in subsequent weeks. This is due to both countries having entries at the 2015 Academy Awards for best Foreign Language Film. Similarly, select Greece and you’ll see strong co-occurrence with Eritrea, reflecting that a number of the immigrants crossing the Mediterranean into Europe come from that country.
This visualization follows in the footsteps of Culturomics 2.0, which four years ago offered the very first mass-scale visualization of the geographic community structure of the world’s news media, followed a year later by one of the early large-scale visualizations of the geographic structure of social media in the Global Twitter Heartbeat. Here however, we leverage GDELT 2.0’s ability to peer deeply into the local presses of the world and to live-translate all of that coverage from 65 languages, coupled with the incredible power of Google’s Arms Globe Visualization, to provide an unprecedentedly high resolution and interactive view of the geographic co-occurrence network of the world’s collective news media output over the last three months.
For the technically-minded, the visualization was produced by processing the complete set of GKG 2.0 files from February 22, 2015 to May 30, 2015 through a set of PERL scripts that extracted and collapsed the fields and output the final JSON data file. The Google BigQuery GKG 2.0 table was used to prototype an early version of the visualization and throughout the final project to spot-check results, making it possible to instantly pull up specific matching coverage and to perform additional summations and confirmations.
Click the link below to launch the visualization in a new browser window!
NOTE: this visualization requires a modern WebGL-capable web browser such as Google Chrome operating on a desktop computer and will NOT work properly on most mobile devices due to the required graphical processing power and minimum screen dimensions.