On Friday, Google published a blog post saying it has updated its machine learning systems to detect and remove fake business reviews on the maps service, as well as fake business listings, photos and videos.
Machine learning systems and human review teams removed more than 200 million photos, 7 million videos, and more than 115 million comments, up 20% from the previous year.
How Google uses machine learning to weed out fake content
1- Recognize patterns of abuse faster:
Google uses new machine learning models to quickly identify and remove fake and fraudulent content. These machine learning models look for unusual patterns in user-submitted content, including: Reporting new forms of abuse that have not occurred before.
“Last year we made significant updates to our machine learning models, which helped us identify new abuse trends much faster than in previous years,” Google said in a post. Detect manually from millions of accounts.
"Our team of analysts quickly confirmed that these pages were fake, and we were able to quickly remove them and disable the accounts associated with them," she added.
It should be noted that the Google system checks new content before it is published to prevent fake or fraudulent content from being submitted to Google Maps and uses machine learning models to analyze previously published content for possible fake content. Pass the pre-test.
2- Delete fraudulent photos:
In some places, scammers put their phone numbers on images they provide to mapping services, hoping to trick users into calling them instead of the actual company. Google used a new machine learning model that is able to compute numbers superimposed on submitted images by analyzing specific visual details and image layouts.
With this form, Google intercepts and blocks the vast majority of these fraudulent images before they are published.
Here are some statistics about the fake content detected by Google:
- Google has blocked more than 115 million fake comments, saying it blocked most of them before they were posted.
- New anti-spam algorithms have removed more than 200 million images and more than 7 million videos that were blurry, low-quality, or otherwise violated Google's content guidelines.
- These new systems prevented 20 million attempts to create fake corporate accounts, an increase of 8 million from 2021.
- Google also has safeguards in place for more than 185,000 businesses that are subject to suspicious activity.