Active Learning

A machine learning paradigm in which the algorithm iteratively selects the most informative unlabeled data points to query a human annotator for labels, enabling effective model training with minimal labeled data. Active learning uses sampling strategies such as uncertainty sampling — selecting instances the model is least confident about — to maximize learning efficiency. In accessibility applications, active learning enables personalized assistive tools that adapt to individual users' needs through minimal interaction, such as learning which words a person who stutters finds difficult to pronounce from just a few examples rather than requiring extensive upfront configuration.