Comment by sbrother
14 hours ago
Seriously. I didn't even realize this was a wide issue, but I couldn't find a school enrolment email I was looking for this morning, and found it in the spam folder. The fact that I basically never have to do this is actually amazing.
I wonder about difference in experience that different people have with gmail’s spam filter. In my case, the majority of emails that go to my gmail spam folder are legitimate. I don’t actually receive much spam, a single-digit number of emails per month (in the past 30 days, 2 emails), so any time I see anything in my spam folder I have to check so that I can rescue the email if legitimate.
This is my experience also. Closely guarded email, haven't received _any_ spam to it to date, but a large volume of false positives. This, among other reasons, actually led to my setting up my own email server again. Gmail is a great product if you don't know what you're doing or have avaliable to you. It's like a McDonalds burger. Not inpressive, not good, bot bad either, and certainly won't offend anyone while being accessible --- but calling it good is a bit out of touch with what good looks like.
I kinda expect there are a lot of false positives that people just never notice because they've got also thousands of unread (non-spam) emails in their inbox and never check their spam to see if there's anything legitimate there.
They probably have a trillion emails with human labels, either from users directly applying them, or inferrable from actions like deleting.
With that much data, even a simple Bayesian classifier should work pretty much perfectly.