Screen Shot 2021-07-07 at 9.52.34 PM

Geospark Analytics

  • Geospark Analytics was founded in July 2017. We set out to solve the problem of information overload: we reduce the time  analysts spend finding relevant information by 30-40 percent.
  • We decided that more data is not better – smarter data is better. Geospark Analytics provides users with smarter data through machine learning and artificial intelligence. Our AI platform, Hyperion, ingests information from news media, social media, travel and weather advisories, economic indicators, and more. It automates intelligence and drives insights in relation to stability and disruption around the world.
  • Geospark Analytics' headquarters is based in Herndon, Virginia. Our clientele have offices and assets worldwide, and we pride ourselves in our platform’s global coverage of open-source data.

  • The Geospark Analytics team consists of a fantastic data science group, veterans of the intel community, and a variety of industry professionals.

  • Hyperion automates data gathering, risk and threat information, and analysis by collecting information from over 6.8 million unique sources, including news, social media, economic indicators, governance indicators, travel warnings, and weather.

  • The AI platform creates workflow efficiencies that save hours per day. We believe the ultimate differentiator is our predictive analytics technology, deployed across the globe into the hands of humans who can verify events and explore additional context specific to their organization within Hyperion.

  • Hyperion's innovative AI-Engine and machine learning models — Events, Hotspots, Pulse, and ScribeAI — give users the information and understanding needed to accomplish the mission.

Hyperion

  • Hyperion is Geospark Analytics' artificial intelligence platform. It intakes 150-200,000 data points per day and geolocates them. 

  • The raw news and social media data is accessible to analysts and can be filtered according to Hyperion features such as Event categories, time frame, and keywords.

  • Hyperion uses news and social media data, in hand with longer-term data (such as biannual regional reports, economic indicators,  et cetera) to generate automatic, quantified stability assessments. This 'stability score' is called Pulse and is a key feature of the Hyperion platform.

Hyperion currently breaks data into the following Event models and sub-models:

  • Travel
  • Unrest
  • Conflict
  • Crime
    • Drugs
    • Kidnappings
    • Shootings
      Trafficking
      More crime
  • Health
  • Terrorism
  • Transportation
    • Aviation
    • Maritime
    • Rail
    • Roadways
    • More Transportation
  • Disasters
    • Earthquakes
    • Evacuations
    • Explosions
    • Fires
    • Floods
    • Gas leaks
    • Hazmat
    • Power Outages
    • Severe Weather
    • More Disasters
  • We do allow for Event model customization, and we have done so in the past for particular clients. You can reach out to your Client Success Manager to begin this discussion.
  • Hyperion can translate content from over 100 different languages.

  • The feeds can also be translated into any language automatically. In the left-hand side bar under settings, users have the ability to change language.

  • Once a new language is set, such as French, all content ingested in French will present in French.

  • When there is an increase in info (news and/or social) outside the norm, the NLP (natural language processor) reads the info, correlates it with our risk Event models, and identifies if the spike is caused by one of the Event models (for example, a protest). If there is enough incoming data outside the norm, a Hotspot will populate on the map.

  • Users can click on the Hotspot to explore the information driving its occurrence. They can also set up a Hotspot alert and receive the content by email and/or push notification.

  • Yes, we do gather information on maritime operations. Typically the information is geolocated to the closest port or land.
  • Yes! We do so through Places, in which we integrate the geocoordinates of a place, its name, and store any other relevant info you want (such as phone number, names, website links, and more).

  • You can read more about how Places can be customized to your needs here.

  • There is currently no way to identify when a source type has been cut off (for example, Nigerian social media). However, often times when there is a ban or blackout of a social media or news outlet in a region or country, there is a lot of negative articles about the ban itself. The platform will take these in.

  • Additionally, it is important to remember how Hyperion geolocates content – not necessarily locating the content geospatially (where the social media posts are published), but rather where the social posts are talking about! This is key to remember, as those social posts that are now identifying the ban in Nigeria are being geolocated to the country (influencing Pulse).

  • You can save any piece of news or social news media data into a content folder. These folders can be shared and collaborated upon among your other Hyperion-using colleagues.

  • You can read more about how to build Content folders and explore the data here.

  • We have created an eLearning portal here.

  • This has a number of user guides and short videos (10 min or less) on all of the major features.

General

  • AI, or Artificial Intelligence, is a branch of computer science that refers to the simulation of human intelligence in machines, achieved through algorithms and applications such as natural language processing (NLP) and self- correction processes.

  • At Geospark Analytics, we use AI in a variety of ways, one being ScribeAI, an automated risk and threat assessment report.

  • ‘Machine learning’ is a branch of AI that automates analytical model building, meaning that systems can learn from data, identify patterns and make decisions with minimal human intervention.

  • Geospark Analytics leverages machine learning to classify news and social media data into different Event models, as well as extract relevant information such as geolocation.

  • Hyperion’s algorithms analyze content and geolocate data to the location of which the content references. To put simply, an article written by the New York times about unrest in Colombia will be located in Colombia, not in New York.

  • Hyperion’s geolocation function tests at 92% accuracy, well-above industry standard for artificial intelligence.

  • Geospark Analytics data scientists are continuously adapting the algorithms to improve. In cases in which data is mislocated, Hyperion users have the capability to mark data as mislocated in the system. This information is fed to our data science team to be reviewed, and help teach the system.

  • It is important to mention that Hyperion functions on a country and regional scale. Its capabilities leverage machine learning and AI models to give users the power of information regarding stability and breaking events. 

     

  • We are not an official misattribution technology. However, there are users that log-in to their own misattribution accounts and use Hyperion in parallel. Within Hyperion your identity is obfuscated, and we can always add in a pop-up to warn users they are leaving the platform and are headed to an original source.