In a bid to curtail offensive memes that promote hate speech, Facebook is building a novel artificial intelligence (AI) system that uses machine Learning How to identify text from images and videos as well as transcribe it.
While resources to transcribe text are nothing new, the organization faces different challenges because of the size, sheer quantity of photos shared every day on Facebook and Instagram, and the number of languages supported on its worldwide platform.
Together with the publication AI system called Rosetta, Facebook can process over a billion public images and Instagram images and video frames (in a huge array of languages) daily through the system economically.
The extracted text is subsequently used by downstream classifiers to instantly act upon policy-violating articles or by product programs like photo search.
“Recognizing text in images along with the context in which it seems helps our systems to identify inappropriate or harmful material and keep our community safe,” the societal networking giant said in a blog article late on Tuesday.
“Text extracted from images is being used to enhance the value and quality of photo search, automatically identifying content which violates our hate-speech coverage on the stage in various languages, and improve the validity of classification of photos in News Feed to surface personalised content,” the media giant noticed.
Further, Facebook is also continuing to invest in expanding the text recognition model for its broad selection of languages used on its own worldwide platform.