By Adam Goodman, CIO-T Strategic Communications and Engagement, and James Brown, Source Foundation Communications
Nearly 1,000 times a day, NGA’s Maritime Safety Watch desk receives messages — usually via satellite or email from governmental authorities or individual ships — about safety of navigation issues.
Pirates. Icebergs. Ship in distress. Man overboard. Missile launches. Rocket launches. Military exercises…. The topics vary, and the number of messages has grown over the years.
The 24x7 Maritime Safety Watch desk is responsible for monitoring those individual reports, for decades manually organizing and validating them and then issuing warning broadcasts — typically about 30 to 40 a day.
Lives can be at stake. Quick and dependable alerts are essential. It is a critical task, and one required under NGA’s Title 10 Safety of Navigation mission.
Now, in a unique collaboration of the agency’s Source Maritime Safety Office, Research and Foundation GEOINT Integrated Program Office, comes a new automated system that promises to speed up and simplify that process to timelier and more accurate navigation warning alerts.
Called Source Maritime Automated Processing System, or SMAPS, the automated system relies on natural language processing and basic machine learning to improve the watch’s largely outdated system.
“Watch officers can now focus more on analysis and turn over the data administration to the software,” explained Chris Janus, branch chief of Maritime Safety Watch, which is part of the Source Maritime Safety Office.
Janus estimates that the automated system will reduce the time it takes to process an incoming message and then send out an alert by up to 50%. Those warnings go to the United States’ Navy, Merchant Marine, Coast Guard and other commercial shipping interests.
In addition, Janus anticipates a significant reduction in data entry and processing errors by moving to an automated process.
One key feature of SMAPS is its ability to structure the incoming data so that watch officers can easily approve, adjust or provide additional feedback, said Mike Lenihan, Research program manager.
Geospatial and category tagging as well as the ability for continued machine learning improvements will position the system for even further future modernization.
“The more info you feed it, the better the model will get,’’ Lenihan said.
Lenihan had previously worked with the Maritime Safety Office in 2019 to develop the Enterprise Engine Notice to Mariners automated processing system, referred to as E2NtM. That system uses artificial intelligence and machine learning to update navigational charts, requiring recognition of text as well as symbols, but on a much less frequent basis.
NGA’s Foundation GEOINT Integrated Program Office will be responsible for sustaining and evolving SMAPS after it has fully transitioned to operations.
“We make sure everything is aligned,’’ said Necole Allen, IPF’s SMAPS program manager. “Once handed off to us, we will sustain the system, make updates and provide maintenance.’’
Lenihan notes that the project exemplifies the importance of inter-office and inter-directorate collaboration within the agency.
“It was a truly coordinated effort,’’ he said.
While maybe not the flashiest of the agency’s artificial-intelligence and automation projects in terms of cutting-edge natural language processing algorithms, it is absolutely a much needed and functional one, with valuable lessons learned to share with the entire agency.
“As the agency leans more toward automating production systems, this is a great first step in achieving that goal,’’ said Lenihan.