Here's how Google wants to fight fake news and offensive content: through new algorithms to protect users
Fighting fake news, online hatred, improving the user experience. Google has put all its power into action to stop the spreading phenomena on the web and make the Internet a better place and stop offensive, violent or misleading online content .
Stop Fake News thanks to research
Ben Gomes, vice president of engineering at Google, told some of the most significant innovations recently put into place by the technological giant in the field of web research. We report the essential steps taken from his blog post, which illustrates the actions taken by Google in the fight against fake news and in order to correctly position the sites in the searches.
Here's what Ben Gomes says about Fake news and offensive content
Web search can always improve. We knew this since I started working on Search in 1999, and it still holds today. At the time, the Internet was expanding at an incredible rate. We had to make sense of the information explosion by organizing and presenting it so that people could find what they were looking for already on the Google results page.
Allepoca revolved around PageRank, the algorithm by which the importance of web pages was measured, to be indexed in the search results. In addition to trying to organize information, our algorithms have always had to face individuals or systems that try to "fool" our systems, in order to appear high in search results, but using low quality content bases, hidden texts and other deceptive practices. We have faced these problems – as well as others – over the years, regularly updating our algorithms and introducing functions that prevent the various deceptions.
Protect the user from Fake News and offensive content
"Today – says Ben Gomes – in a world where tens of thousands of web pages are born per minute, there are always new ways through which people try to deceive systems. The most relevant phenomenon is that of Fake News, the false news, whose content has contributed to spreading falsehoods, offenses and manipulated information. Although this problem is very different from the one I mentioned before, our goal remains the same: to allow people to access the most relevant information through the reliable sources available "." And although we cannot always do it at best, we are however making good progress in addressing this problem, but in order to have long-term changes, impacting changes, structural research is needed "
High quality against false news
"In terms of long-term interventions – continues Ben Gomes of Google – today we are moving forward high quality in web search. It includes improvements in the ranking of results, easier ways to provide direct answers to people and also greater transparency in web search".
Algorithms to verify reliable sources
According to Bem Gomes "Our people algorithms allow us to identify the sources reliable among the hundreds of billions of pages in our index. However it is clear that a small part of ours traffic daily (around 0.25%) appears to be made of content offensive or misleading, which does not correspond to what people look for. To avoid the diffusion of this content we have improved our evaluation methods and have made algorithmic updates to give priority to the most authoritative contents. New Search Quality Rater provide for a trial process. As part of this process, they will be employed assessments from real people who will value the quality of Google search results by giving feedback on our tests. These assessments will then be used to help us collect data on the quality of our results and to identify areas where we need to improve. " In short, real people will read and try to understand the meaning of fake news, improving the results and deleting what is deceptive.
Search Quality Rater
"Last month – recalls Ben Gomes of Google – we updated the guidelines of our Search Quality Rater to provide examples of fake news and low quality web pages that could include misleading information, unexpected offensive results, scams, theories and unfounded conspiracies. . These guidelines will begin to help our algorithms reduce poor quality content from search results. The changes in the rankinginstead, they made it possible to combine hundreds of signals to determine what results were shown based on a specific query. Judgment criteria include: the date of content creation and the number of times search queries are displayed on a particular page ".
Autocomplete without offensive content
Gomes continues: “When using Google, we propose to accelerate the user experience with features such as lAutocomplete, the automatic completion, which helps to predict the searches you are typing so as to quickly obtain the necessary information, through the snippets in the foreground the most relevant information of the research will be shown. The content displayed by these features will be algorithmically generated and will reflect what people search on the web. This can sometimes lead to unexpected, inaccurate or even offensive results. Starting today, however, we have facilitated the possibility of mark a content displayed. These new feedback mechanisms include categories that have been clearly labeled so you can directly inform them if you come across sensitive or unnecessary content. And we're going to use this feedback to improve our algorithms ".
Users shocked by Google search
"In recent months, we have been asked difficult questions about forecasts shocking or offensive that appear in autocomplete – admits Ben Gomes -. Based on this, we evaluated where to improve our content criteria and how to update them appropriately. We are now publishing this new practice to the Help Center, so that everyone can know more about Autocomplete and our approach to removals ”.
How Google search works
“For those wishing to learn more, we have recently updated our website How Search Works to provide even more information to users and website owners about the technology behind the search. The site includes a description of how Google ranking systems order hundreds of billions of pages to offer both results and an overview of our user testing process. "" Each year there are trillions of searches on Google and 15% of the searches we see every day are new, which means that there is always more work to be done to offer people the best answers to their questions, obtainable from a wide variety of legitimate sources. Although our web search results will never be perfect, we are committed – as always – to maintaining trust and ensuring that our products continue to be useful to everyone.