It's an interesting business to be in these days...
When, Why and how did you get into it?
Does it generate meaningful/worthwhile revenue or is to more to cover overhead? (I only noticed the occasional 30 second commercial intro) Maybe you have a lot of premium subs?
Why sometimes do certain cities/counties change their feed? For example one night the Minneapolis police were on a completely different county that was hours away.
Majority of the time, these feeds are relatively low key right...with actual concerning/escalated incidents few and far between. Wondering: is there someway to us ML to identify specific things and provide push notifications to people in a given geography around that? For example it could listen for "shots fired" in a specific area and notify me via sms or whatnot when that occurs.
And AI could be used to show more of an abstract map view if and when "violence" is rising based on action on the scanner, right?
The business is wildly successful. Revenue is a split between paying subscribers, license and royalty revenue, and advertising.
About 90% of the feeds are provided by volunteers who simply connect a radio to their computer and broadcast to us. The other 10% are actually provided by the agencies themselves. We probably turn over 20 feeds a day, so coverage comes and goes all the time. There is a lot of work going on in the ML / AI space around the content, but it is a hard problem to solve because the quality of the content diverges wildly. Vernacular differs. Coverage comes and goes. Organizations like Citizen, which have HUGE funding (60 MM lol!), are trying to solve this, but they still end up just employing armies of workers to sift through the data and audio and normalize it all.
We're doing a lot of innovation in the space on our end, including using SDRs to vacuum up wide swaths of RF and then store call data, which is going to be the next "revolution" in our space. Interesting times...
Been there, done that. The technique is scary-good, even with cheap hardware.
Capture everything, sort by talkgroups, form listen queues ("police", "fire", "police NOT university",) prioritize, and place recordings in queues. Stream queues through ice cast. Drop low-pri messages if you fall too far behind real-time.
Speaking of: if anyone wants nearly two years of uninterrupted Seattle police radio archives (and everything else from KCERS), they should get in touch.