Google has reduced spam reaching inboxes to a fraction of a percent, but in the process sometimes misclassifies bulk-mailed messages like monthly statements and ticket receipts.
It’s a big problem for large bulk emailers of legitimate messages. To deal with it, Google has created a toolset to help those mailers figure out what’s happening to their messages.
Postmaster Tools is designed for administrators and provides information on delivery errors, spam reports and reputation, wrote Sri Harsha Somanchi, a Google product manager, in a blog post.
Google is also try to make its spam filter for consumer accounts more customizable. Although people can already classify messages by specifically labeling one as spam, what is spam to one person may be considered a desired communication by another.
“So while your neighbor may love weekly email newsletters, you may loathe them,” Somanchi wrote. “With advances in machine learning, the spam filter can now reflect these individual preferences.”
The company’s spam filter is now using what it calls an artificial neural network, which involves a large distributed computing platform that use self-training algorithms. Somanchi wrote that it can sort out sneaky spam, which looks very close to desired messages.
That platform can also figure out spoofed messages, which appear to come from a legitimate sender but actually do not, Somanchi wrote.