Amazon Improves Reviews System with Machine Learning

In its ongoing effort to improve the value of customer reviews -- and reduce the likelihood of fake, paid-for reviews -- Amazon has enhanced its system with machine learning. Being rolled out just in the U.S. for now, the improved rating system is designed to "learn which reviews are most helpful to customers" and continue to improve over time.

The updated system comes on the heels of a lawsuit filed by Amazon in April against an alleged purveyor of fake reviews. That suit alleges that for-profit, misleading reviews undermine customer trust in Amazon and tarnish its brand.

Claiming to have pioneered the use of online customer reviews 20 years ago, Amazon said its site now hosts "hundreds of millions of unique reviews." Although its user policies prohibit false and manipulated reviews, "an unhealthy ecosystem is developing outside of Amazon to supply inauthentic reviews," according to the company.

Newer Reviews To Gain the Edge

Julie Law, a spokesperson for Amazon, told us that the new system "will use a machine-learned model to give more weight to newer, more helpful reviews from Amazon customers. The system will continue to learn which reviews are most helpful to customers and will improve the experience over time."

The enhancements will affect both star ratings and review rankings, Law said. The star rating system "will now consider factors including the age of a review, helpful votes by customers, and whether the reviews are from verified purchasers," she said.

Machine learning will also be used to determine where a review is ranked in the list of reviews for each product. "We hope these changes will help customers make even more informed purchasing decisions," Law said.

Legal Challenge Targets Misleading Reviews

"Amazon has a long legacy in machine learning," Jeff Bilger, Senior Manager of Amazon Machine Learning, said in April when Amazon Web...

Comments are closed.