Taking on good chances as fake news may reside inside of photos and videos, Facebook is expanding their photos and video fact-checking program to all of its third-party information checking organizations.
Facebook intended to combat fake news through this initiative that can propel misinformation campaigns in news articles to generate ad revenue.
That said, the company has to be capable to tackle other forms of fake news that may be residing inside of images and the audio that accompanies video clips.
First time in March this year, Facebook detailing photo and video fact-checking project has said that the Agence France-Presse (AFP) news service would participate.
Now, announcing the expansion, the social networking company said that all of its “27 fact-checking partners in 17 countries” would now have access to and contribute to the video and photo screening technology.
This will help them to identify and take action against more types of misinformation.
Detailing about the work process, the company said, a Machine Learning (ML) based model by using various engagement signals, including people’s feedback to determine whether a particular photo or video has been doctored or deemed “potentially false content.”
Facebook’s third-party partners, then verify and rate the images or videos, for further scrutiny. These independent fact-checkers can also surface content on their own for review and rating.
A video below shows how the machine learning model detects false photos and videos:
“Many of our third-party fact-checking partners have expertise evaluating photos and videos and are trained in visual verification techniques, such as a reverse image searching and analyzing image metadata, like when and where the photo or video was taken,” Facebook wrote in a statement.
“Fact-checkers are able to assess the truth or falsity of a photo or video by combining these skills with other journalistic practices, like using research from experts, academics or government agencies,” the company said.
The company says it hopes to improve the accuracy of this machine-learning model with more ratings from these outside partners on photos and videos.
Facebook is also leveraging optical character recognition (OCR) technology to extract text from photos and then compare this text to headlines from fact-checkers’ articles.
Even more, it’s working on more new ways to detect the manipulated photos and videos. So, they can identify and send more potentially deceptive photos and videos to the fact-checkers for manual review.
Facebook has categorized false photos and videos into three categories, including “Manipulated or Fabricated,” “Out of Context,” and “Text or Audio Claim.”
Here is an image of these three types of categories of false photos and videos:
Lastly, the company acknowledges that fighting false news is a long-term commitment as the bad actors are always changing their tactics. And, will continue to invest in more technology and partnerships.