A Vision Of The Future: GDELT + Google Cloud Vision API

The past year has been an incredible year for GDELT. With the public release of GDELT 2.0, we now monitor one of the most extensive catalogs of local media in the world, translate absolutely everything we monitor in 65 languages using one of the largest deployments of streaming machine translation, assess up to 4.5 billion emotional indicators per day through one of the largest deployments of sentiment analysis, and geocode more than a billion location references through one of the largest deployments of multilingual fulltext geocoding, to name just a few. Yet, a commonality in all of this is that to date GDELT has only processed the text of global news coverage – the massive firehose of imagery that accompanies worldwide news coverage has been cataloged by GDELT, but GDELT has never been able to offer insights into the contents of all of that imagery.

We are therefore enormously excited to announce that GDELT is one of the early release users of the new Google Cloud Vision API announced this past Wednesday. Look for a number of announcements over the coming weeks as we begin to integrate the Cloud Vision API into the various GDELT pipelines and slowly ramp up its deployment. Using the Cloud Vision API we are beginning to process the primary imagery featured in each news article we monitor, generating a list of objects, actions, and topics found in the image, recognizing any text found on store fronts, boxes, vehicles, and other objects in the scene, and even performing image-based visual sentiment analysis, identifying whether people in the image are happy or sad, angry or surprised. In just our first 24 hours of working with the new API we've been astounded by how accurate and robust it is across the massively diverse and globalized imagery that GDELT observes from across the planet.

To our knowledge this integration marks the first-ever global-scale deployment of deep learning neural network image recognition on worldwide news imagery from every corner of the earth and every topic imaginable, and marks the final milestone in allowing GDELT to more fully make sense of the world's news media as a human analyst might, comprehending not only the text on the page in 65 languages, but the visual imagery that accompanies it.

Stay tuned for this incredible new vision of the future!