Machine Learning Helps Google Reduce Gmail Spam

The machines might not ultimately win in "Terminator" movies, but they are prevailing in the neverending battle against spam. In fact, machine learning has helped Google reduce unwanted e-mails in Gmail inboxes to less than 0.1 percent.

In the meantime, to help further reduce the odds of wanted e-mail being misidentified as spam, the search giant on Thursday launched new Gmail Postmaster Tools for legitimate, high-volume e-mail senders like banks and airlines.

Currently, only 0.05 percent of wanted e-mails end up in Gmail users' spam folders, according to Google. However, better analysis and artificial intelligence can help reduce that figure even more, avoiding the need for people to go "Dumpster diving" for important e-mails buried among the spam, Google said.

The new Gmail Postmaster Tools are designed to help companies that send out large volumes of e-mails to customers "analyze your e-mail performance and help Gmail route it to the right place," according to Google. The tools provide qualified e-mail administrators with Google data on delivery errors, spam reports, reputation and more to help them fine-tune their e-mail delivery.

Goal: 'Spam-Free Gmail'

Using, among other things, the data provided by Gmail users when they click the "Report spam" or "Not spam" buttons, Google's machine learning applications have enabled the company to steadily reduce the amount of unwanted e-mails landing in users' inboxes.

"When you click the 'Report spam' and 'Not spam' buttons, you're not only improving your Gmail experience right then and there, you're also training Gmail's filters to identify spam vs. wanted mail in the future," Gmail Product Manager Sri Harsha Somanchi said in a post on the Official Gmail Blog. "Ultimately, we aspire to a spam-free Gmail experience."

For example, Gmail's spam filter now uses an artificial neural network "to detect and block the especially sneaky spam -- the kind...

Comments are closed.