Over 4 Million Fake Business Profiles and Other Policy-Violating Data Was Removed By Google Maps Using AI and Human Moderation

On a daily basis, millions of people give around 20 million reviews, ratings and other feedback on Google Maps’ 200 million points of interest. The platform is ever-growing but these contributions are 100% perfect. Therefore, Google tries to keep the service secure by identifying the malicious contributions using artificial intelligence (AI) and machine learning at the time of submission so it does not catches the eye of over 1 billion daily users.

Google uses automated detection systems and machine learning models to scan millions of contributions. If policy-violating content is identified, it is automatically removed from the platform. When a review is posted, Google’s system audit each review, ensuring it does not contain bogus or misleading content, before it is published on Maps.

The machine learning model looks for certain words and phrases and examines the pattern in which the user has previously contributed and checks if there is any apprehensive review pattern.

AI cannot be always trusted, therefore Google has a dedicated team of trained operators and analysts to audit reviews, pictures, business profiles, and other content.

According to Kevin Reece Director of Product at Google Maps, in 2019, machine learning systems helped in removing the policy-violating content and irregularities were detected for manual reviews, resulting in the removal of 75 million policy-violation reviews and 4 million false business profiles through human moderators. Around 580,000 reviews and 258,000 reported business profiles were also removed. Plus, over 10 million photos and 3 million videos were also reviewed and removed manually for violating the policies on Maps. More than 475,000 users were also disabled for repeatedly violating the policies and guidelines.

Kevin Reece further explained in a blog post that a majority of contributions of Maps are authentic and only less than 1 percent content was identified violating policies. The contributed content is indeed an important part of Maps, which makes it helpful for all.

The action was taken after millions of fake listing on Maps were reported by media outlets, which was then removed by the moderation team of Maps. There have been various controversies over the user-contributed content after which Google had to take down the public map editing tools and bring new features for Maps moderation.

Google is not the only platform to use AI and machine learning for moderation and identify misleading content. Last year in October, Pinterest reported that around 88 percent of the self-harm content was reduced after implementing the AI. At the same time, Twitter also came forward, claiming 50 percent of the abusive tweets are flagged automatically before users report it. Also, more than 95 percent of the indecent content that violates the policies of Facebook was identified through AI.

According to Google, Maps includes over 220 countries and there are around 120 million members of the community that contributes reviews and other content on Maps, known as the Local Guides.

Maps include the accessibility info like the wheelchair-friendly entrances of more than 50 million places worldwide, and weekly over 5 million websites and apps with Maps integrated into them use the platform as well.



Read next: Google introduces AI-powered tool AutoFlip to auto-adjust video dimension

No comments:

Post a Comment