February 22, 2018

Using Our Real-Time Knowledge Engine to Alert on Earthquakes Faster Than Government Sources

Sean Solbak by Sean Solbak

Two weeks ago, we unveiled the public beta of our real-time knowledge engine. Here is one example of how we help our users know what is happening, when it happens.

At 14:03 GMT, on Saturday Feb. 17th, a 4.6 magnitude earthquake was registered 20 km northeast of Swansea, Wales, UK. The first public acknowledgement of the earthquake came from the British Geological Survey (BGS) 50 minutes later, stating they were analysing the data. 1 hour and 13 minutes after the quake, BGS announced that Swansea was the nearest and hardest hit.

While SAM does pull in critical official reports, our system is not solely reliant on government data or media channels. SAM’s AI scans social chatter from on-the-ground sources to detect anomalies and triangulate where these events are happening. Our real-time knowledge engine detected the first Swansea earthquake report at 14:33 GMT and alerted our users at 14:35 GMT. This gave our users an 18 minute head start over the first BGS alert and a 39 minute head start compared to the first BBC Breaking alert.

SAM’s approach is global, so it doesn’t rely on any one government or official source to power our event detection. Which is extremely important as some governments do not openly share critical data or have the necessary infrastructure to do so—making it hard to know what is happening on the ground in a timely or reliable way. We saw evidence of this a few days after the Swansea event, when a small earthquake hit Kathmandu, putting the city on edge. In this case, SAM first detected the Kathmandu earthquake at 20:38 GMT and alerted a mere 60 seconds after the quake itself, giving a 43 minute lead time over any credible media source.

Without any human intervention, SAM’s AI understands the meaning of social media posts, scores the uploader’s credibility, and finally works to place these posts in space and time. In many cases, when we detect an event, the initial reports are from everyday people whose social posts on their own would not be enough to warrant a credible alert to our users. Depending on the authenticity/credibility of the person making the social media report, SAM will work to find additional social sources who can corroborate the event. Or debunk, in the case that spam bots or dubious social media accounts attempt to create a fake spike in social post volume.

You can think of SAM’s real-time knowledge engine as your very own AI Analyst who can distill world events into a credible stream of information. In both of the cases above, SAM’s users were able to quickly ensure the safety of their people, know it was still safe to land an aircraft, and monitor potential escalation for any required emergency deployments.

By continuing to use this site you consent to the use of cookies in accordance with our cookie policy. Learn more