Thu Dec 02, 2021 2:46 pm
#1886097
Tim Dawson wrote:The trouble with that, and any setting I implemented that let you hide all NOTAM relating to obstacles being erected, is that I start imagining the accident report. Something about the software the pilot was using allowing them to suppress notifications of obstacles being erected. It doesn't look good.
I recreated your route around London - though it wasn't clear which airfield was your takeoff and landing point - and managed to get 4 such obstacle NOTAMs in my briefing. That's a lot less than a NATS AIS brief would include because we skip NOTAMs that you'll be flying above. Also by tightening up the horizontal distance parameter on my narrow route brief to 3nm I've just managed to cut that number in half, and the ones remaining are at my planned takeoff airfield.
A much bigger problem is all the notices about overflying far-flung countries that I'm not planning to go anywhere near, and covid stuff. There were 8 long bulletins like that for the same route. Also a page-long NOTAM about the B737. Unfortunately the system lacks any flag that we could rely on to filter out such things.
I'm absolutely not suggesting it's your remit to look at this, but I think about Waze and the wisdom of crowds and wonder if there's a piece of discovery that could be done using AWS Comprehend or similar, to see if NOTAMs could be machine classified in some way based on common phrases.
Then - to bring the Waze example into it - I wonder if you could leverage other SD users' behaviour to see which NOTAMs people read, which ones they always close etc - and rank/score/augment them
The output from this wouldn't be to hide or suppress the NOTAMs, but maybe you could order them in some more meaningful way: most impactful >> least impactful, or colour-code them according to user feedback etc. Certain words and phrases 'searchlight, fireworks, laser, balloon etc' should enable fairly positive classification. Ditto, the word 'Belarus' or 'Iraq'.
I get it's different for different people, but natural language processing is pretty powerful and would be a nice hack project.