Gmail's spam filter is enhanced with a neural network, and new tools for mass senders are also released

ByWill Lion

Gmail's spam filter has succeeded in reducing the probability of spam mail entering the inbox on average 0.1% or less, and the probability of classifying e-mails to be read as junk mail folders to 0.05% or less. Gmail's spam filter that already has such a high classification function is the same as Google search and Google Nowneural networkTo apply classification with higher precision, and furthermore, a sender's tool "not to be classified as spam mail" which is easy to be classified as spam mail such as monthly statement and receipt "Postmaster Tools"Has also been made public.

Postmaster Tools - Google
https://gmail.com/postmaster/


Official Gmail Blog: The mail you want, not the spam you do not
http://gmailblog.blogspot.jp/2015/07/the-mail-you-want-not-spam-you-dont.html


According to the Gmail blog, Gmail's spam filter realizes high precision even at present, but it is classified as spam, such as a newsletter that you want to read or some confirmation code that arrives when you register for a service There is that. If the mail you want to read is classified as a spam folder, the user will be looking for the necessary mail out of a large number of unnecessary mails, so the Gmail team will make this error as low as possible We released a new tool of.

Postmaster Tools for senders is a dashboard for people who send large numbers of e-mails in business and so on, can register domains to use, analyze outgoing mail, reputation of their domain, send You can diagnose the percentage of mails being reported as spam and the status of delivery errors. Based on these diagnostic results, the sender can improve the sent mail so that it is not considered spam.


Also, some people like newsletters that are treated as spam are difficult to classify, but by installing the same neural network as Google search and Google Now on the spam filter, "Report spam" "Spam mail You can learn the behavior of the user "Release" to classify e-mails required for individual users into the inbox. Detection functions such as camouflaged mail and phishing mail will be improved by machine learning.

The Gmail team is ultimately aiming to completely eliminate spam mails from Gmail, and we will continue to exclude spam mails while classifying the necessary mails in the inbox.