[ad_1]
Facebook moderators can not look through every image published on the huge platform. Facebook builds an AI to help them. In a blog post published today, Facebook describes a system called Rosetta that uses machine learning to identify the text of images and videos, and then transcribes it into a machine-readable file. In particular, Facebook finds this tool useful for transcribing text on the same.
Text transcription tools are not new, but Facebook faces different challenges because of the size of its platform and the variety of images it sees. Rosetta is supposed to be online right now, by extracting text from 1 billion images and video images a day on Facebook and Instagram.
At the moment, Facebook does not know exactly what Facebook does with the data. It's useful for basic features like looking for photos and screen readers, says the post. But it also seems that Facebook is starting to achieve much more important goals, like figuring out what would be interesting to put in your newsfeed and, more importantly, determining which memes are just mocking memes and spreading hate speech . or other offensive comments.
Facebook says that text and machine learning are used to "automatically identify content that violates our hate policy" and that it does so in multiple languages. Given the moderation problems well known to society, a system that works well and that can automatically flag potentially problematic images could be a real help.
Source link
