Yesterday, Google wrote in a blog post that the company is using Artificial Intelligence and machine learning techniques to more quickly recognize breaking news around various crises such as natural disasters. According to Google’s VP of search Pandu Nayak, the AI systems of Google now take minutes to recognize breaking stories. In comparison, the detection time of its systems was up to 40 minutes only a few years ago. Nayak wrote in the post that the company has improved its system to automatically recognize breaking news around crisis moments and make sure that the company is returning the most authentic information available.
Likely, quicker breaking news detection will become more critical as the 2020 United States Presidential Election day nears and natural disasters across the globe unfold. Wildfires can change course on a dime, and accurate and timely election information in the face of misleading information campaigns is probably the key to protect the integrity of the processes.
Google’s Vice President of search also wrote that as news is developing, the freshest information published is not always the most authentic and accurate. The company’s improvements in recognizing crisis expand to Google’s work back in 2017 to enhance the quality of results for topics that might be susceptible to offensive and misinformation.
Google conducts extensive testing to make sure that Google Search is as helpful as it can be. The company has done over 1 million search quality tests since 2017, and now the company performs over 1,000 search quality tests each day.
Recently, the company launched an update using language understanding models based on BERT to enhance the matching between news and fact checks that are available. The company started adding publisher fact checks of public claims alongside the results back in April 2017, and Nayak wrote that Google’s systems can now better understand if a fact check of public claim is related to the story’s topic. Its systems surface those fact checks more prominently in the Full Coverage feature of Google News.
Knowledge Panels in Search results display snippets from Wikipedia which is one of the sources fueling the Knowledge Graph of Google. Pandu Nayak said that Google’s machine learning tools are now able to better prevent potentially false information from showing up in search results. When inaccurate information from vandalized Wikipedia pages slips through, Google’s systems are now able to detect such cases with 99% accuracy, Nayak claimed.