Identifying quality of labeled data

This invention helps you figure out how good the labels are on a big collection of information. Imagine you have a huge photo album where every picture is tagged, but you're not sure if all the tags are correct. This method takes a small, detailed peek at some of those pictures, re-tags them perfectly, and then compares those perfect tags to the original ones to give you a quality score. If the score is low, it tells you the whole album might need better tagging, especially if you're using those tags for a specific AI program.