Google’s biggest problem at present is the quality of its search engine results. Granted, this is not a new issue, but it has gotten more publicity recently. Part of the problem is the sudden growth in fake news, especially during the last U.S. presidential election. As part of Project Owl, Google has been using search quality raters in its efforts to combat fake news. However, it is not practical to expect human raters, even in the thousands, to cover the most common search terms. This is where machine learning comes in.
Google employs teams of search quality raters to go through the content listed in search results for a keyword and then rate them in several ways. One of the ways is the quality of the content, and another is if the content is offensive in some way. Google will then use machine learning to teach it’s AI to be able to identify problematic content as well as rate results.
If the effort to train a neural network to identify offensive content works, then Google will be able to use machine learning to cover its entire index and identify and demote problematic content. Little is known about operational details, but it is understood that raters have some form of a guide on how to approach certain forms of content.
This is a good start to improving search quality as well as reducing the prominence of fake news in search results.