Definitely easier said than done, but the process at least makes labelling interesting. Sometimes you run into roadblocks where you can't get past just having a

Definitely easier said than done, but the process at least makes labelling interesting. Sometimes you run into roadblocks where you can't get past just having a human doing some element of the labelling, but once you do have a few algorithmic strategies that work reasonably well on a representative sample of your data, you can usually scale them pretty effectively to the rest of your data