“TensorFlow is Google’s machine learning tool that not only helps blocking trickier mails but also understands contextually different forms of spamming”
Spam is a big issue for all of our inboxes. While some are seemingly harmless (such as the newsletter to a journal that you no longer read), others often target you with critical malware and phishing links. Having dealt with these issues for so long, Google has seemingly become quite efficient at filtering out spam messages, in a bid to ensure that no messages, which you do not wish to see, reach your inbox. Yet, there is a trickier bunch of spammers and spam emails, which find workarounds to Google’s filters, and still manages to reach you.
It is to tackle this that Gmail has started using TensorFlow, its machine learning tool, to better understand what comprises spam emails that its filters have not managed to resolve yet. Using this, Gmail claims, helps it block and filter 100 million more spam emails every day, over the previously existing spam filters. The world’s largest email vendor was already using a proprietary machine learning algorithm, along with a rule and exception-based mechanism, to filter spam mails. The advent of TensorFlow, however, improves spam filtering performance even further.
The powerful machine learning tool also enables contextual understanding of what constitutes spam to different individuals. For instance, even though the content may be apt, regular emails from a journal that you no longer read would constitute as spam to you, albeit not in the conventional sense. Along with this, TensorFlow also helps trace spam mails that do not get registered with Google’s usual filters.
Explaining this in more detail, Google said, “ML-based protections help us make granular decisions based on many different factors. Just because some of an email’s characteristics match up to those commonly considered “spammy,” doesn’t necessarily mean it’s spam. ML allows us to look at all of these signals together to make a determination.“
Gmail is also looking to train new models evaluate pipelines in the machine learning scheme, with TensorFlow. This, Google hopes, would further improve a platform that is already in use by everyone.