We are enormously excited to announce today the official launch of the GDELT Live realtime global dashboard! With the release of GDELT 2.0, we've received an absolutely enormous outpouring of interest in being able to use GDELT to track emerging situations around the world without having to write code or work directly with the raw GDELT CSV files. From NGOs to governments, humanitarian and relief workers to ordinary citizens, you've asked us for a way to track the earliest glimmers of emerging situations across the globe, looking across local media in the most remote corners of the globe in 65 languages in realtime and summarizing that in an easy-to-use intuitive and interactive dashboard. With the debut today of GDELT Live we believe we've answered that call in spades.
GDELT Live is designed to provide a simple easy-to-use dashboard interface to a small subset of GDELT's enormous power, making it possible to use GDELT to track evolving situations around the world without writing a line of code. You can type in any combination of people, organizations, locations, and recognized themes (there are over 2,500 now) into its search box and within a few seconds GDELT Live will show you a live summary of everything GDELT has seen around the world over the last 24 hours about your query, updating every 15 minutes to give you a near-realtime view of the planet.
To help you find your way through the enormous richness that is GDELT's metadata catalog of the planet, the GDELT Live search box autocompletes with the names, locations, and themes that are trending in the last 15 minutes, making quick work of finding interesting topics to explore and helping you align your query with the themes GDELT monitors. It is important to note that GDELT Live is NOT a fulltext news search engine like Google News. You can't type an arbitrary keyword or phrase into its search box. Instead, GDELT identifies the people, organizations, locations, and themes defining the planet's news media in realtime. GDELT may not recognize every name in every language and currently only recognizes a predefined list of around 2,500 themes, ranging from the World Bank Topical Taxonomy to infectious diseases to social, political, religious, ethnic, and terror groups, to animals and plant life, and not all themes may be recognized as robustly in every language.
The Dashboard
Once you've run your search, the Timeline shows the total volume of matching coverage over the last 24 hours in 15 minute increments, making it easy to see when an emerging situation first hit the world's news and whether it is surging in coverage or fading from view. A set of "At-A-Glance" tables summarize the top themes, people, person types (roles like "president", "day laborer", or "child"), general types (religions, ethnicities, political groups, terror groups, weapons, diseases, etc), and locations. Quoted Statements offers a list of trending soundbites and quoted statements from around the world found in matching coverage, while Counts & Amounts lists trending numeric counts, offering numeric context to the size, scale, and impact of your topic of interest. Visual Coverage shows the trending imagery being used to visually represent your topic of interest, while All Coverage provides an exhaustive list of all matching coverage by language. Geographic Heatmap offers a clickable and zoomable heatmap of the worldwide locations most commonly mentioned in matching coverage, visualizing its geographic affinity. Finally, the Thematic Network, Person Network, and Geographic Network offer clickable and zoomable interactive network diagrams showing how the themes, people, and locations of the world are connected within the matching coverage and the key central entities around which everything else revolves.
Enormous effort has gone into making this dashboard fully interactive, responsive, and mobile and touch-optimized. Each of the visualizations and displays automatically adapt to your screen and on mobile devices self-optimize to fit on your screen and make full use of touch gestures and commands. The dashboard has been tested on a wide array of tablet and smartphone devices, but the advanced features may not work on every device and require a modern browser to enjoy full capability.
Dashboard In Action: India and Yemen
To demonstrate the kinds of realtime insights onto emerging world events made possible through GDELT Live, two major world events from today illustrate the dashboard in action and showcase the enormous power it brings to bear looking across languages and geography in tracking a huge cross-section of available global coverage about a situation, surfacing the known details and evolving reaction to it from around the world.
In India this morning a train jumped the tracks near Lucknow, leaving at least 30 dead and 50 injured. As of 9:30AM EST this morning, a search on Google News for the phrase “india derailment” yielded just 71 articles in total and did not offer a “Realtime Coverage” stream for the story, indicating that its algorithms had not detected the derailment as a major emerging news story. In contrast, by that same time, GDELT Live had monitored 615 articles from 25 languages: 400 English, 35 Hindi, 35 Spanish, 24 French, 17 Swedish, 15 Portuguese, 13 Turkish, 13 German, 9 Vietnamese, 8 Romanian, 6 Italian, 6 Russian, 6 Slovak, 6 Bengali, 5 Malayalam, 4 Chinese, 3 Greek, 2 Marathi, 2 Czech, 1 Kannada, 1 Bulgarian, 1 Galician, 1 Telugu, 1 Gujarti, and 1 Ukrainian article. GDELT picked up the first mentions of the derailment between 1:45AM and 2AM EST, with a steady linear surge in coverage through 9:30AM EST as seen in the timeline above (the graph is in 15 minute increments).
GDELT’s realtime machine translation ensures that every article it monitors in 65 languages are translated in realtime as they are monitored, so it is able to transparently reach across all 65 languages with each search. To replicate this level of multi-lingual coverage in Google News would require translating “India” and all of the various common words and phrases for a train derailment into all 65 languages and conduct an individual keyword search for each of them, requiring several hundred searches, and with no ability to look across the coverage holistically for major trends.
The “Counts & Amounts” field is particularly critical for breaking situations like this in surfacing consensus numeric magnitude information for the disaster. As of 9:30AM EST, top counts for the derailment included “2 coaches of the Dehradun”, “3 coaches of the train”, “2 mangled carriages”, “50 injured”, “31 bodies”, “2 children were killed”, “20 ambulances were sent”, and “150 beds have been reserved”. These give an immediate sense of the size of the disaster and response.
In Yemen today, four suicide bombers struck two mosques in the capital of Sanaa, killing upwards of 130 and wounding more than 300. As of 9:30AM EST this morning, a search on Google News for “yemen bombing” or “yemen suicide” yielded just 56 articles. A blue “Realtime Coverage” button was available offering 400 additional articles, but browsing through the results, only the first 50 were actually related to today’s attack, while the remainder were links to past suicide bombings in Yemen. In contrast, GDELT by that point had monitored 978 articles from 33 languages including Arabic, Bengali, Bosnian, Bulgarian, Chinese, Croatian, Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Lithuathian, Macedonian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Spanish, Swedish, Tamil, Turkish, and Vietnamese offering an incredible number of perspectives, imagery, and details of the attacks.
The raw and chaotic nature of news reporting on breaking events is clearly visible in the Counts & Amounts values for the attacks. Top counts centered on “2 mosques” and 2 to 4 “suicide bombers”, but the number of dead and wounded varied substantially and included “13 dead”, “15 killed”, “25 dead”, “40 killed”, “45 killed”, "46 murdered", "50 slain", and "100 injured." By noon EST, however, the numbers had stabilized at “345 injured” and “137 dead”. This ability to constantly monitor death and injury counts in near-realtime every 15 minutes as more and more details emerge from the scene of a disaster offers enormous capability in triaging situations through hundreds or even thousands of sources.
GDELT picked up the first surge in mentions of the attack around 6:45AM EST, while the first article about the attack did not appear on Wikipedia until 8:54AM EST and was merely a one-sentence placeholder that a bombing had taken place at a mosque in Sanaa. It was not until around 10AM EST that the first real details were published to the Wikipedia entry, more than 3 hours after GDELT detected the initial first surge in coverage of the terror attacks.
GDELT Live API
Finally, for those wishing to use GDELT Live as an API for integrating GDELT 2.0 data into other applications, we've added a few features just for you. After running a search, three links are displayed in the top-right of the search results page. The Article List link simply returns a CSV spreadsheet containing an exhaustive list of all of the URLs of coverage that GDELT monitored matching your query. This makes it easy to forward this list on for human review. The GeoJSON link compiles the geographic data behind the heatmap into the GeoJSON file format, suitable for importing into any major online mapping platform. Requesting this link every 15 minutes offers essentially a live-updating geographic stream of your topic of interest. Finally, the Raw GKG link returns the complete GKG 2.0 CSV records from the last 15 minutes that matched your query. Requesting this URL every 15 minutes offers a simple filtering API to stream a filtered version of the GKG 2.0 into your application.
We are incredibly excited about the potential of GDELT Live and can't wait to see what you are able to do with it!