ARCHIVES

The rapid growth of educational video content on platforms such as YouTube has created a significant challenge for learners: transforming passive video consumption into measurable active learning outcomes. Existing Learning Management Systems (LMSs) primarily rely on fixed assessments that do not account for variations in the semantic complexity of instructional videos. This paper presents LearnFull, an AI-driven adaptive learning framework designed to address this limitation by dynamically generating personalized learning artifacts based on the semantic complexity of educational video content. Learn Full employs a Semantic Audit Pipeline powered by the Gemini 1.5 Flash Large Language Model (LLM) to process video transcripts in real time. The system calculates an Information Density (ID) score on a normalized scale of 1.0 to 10.0, which is used to dynamically adjust both the quantity and cognitive complexity of the generated learning artifacts in accordance with Bloom's Revised Taxonomy. To further enhance the learning experience, the platform integrates a Judge0-based sandboxed code execution environment and a gamification system built on PostgreSQL and Supabase, promoting active engagement and supporting programming education. We did some testing to see how well people learn with the version of LearnFull. It looks like people who used this version were 25 percent more likely to finish the technical videos and get a good score. Above 7.0. This tells us that the new way of teaching, which is like a helper that knows how hard something is, really works. People are more interested. They remember more of what they learned. Learn Full is using Artificial Intelligence in Education which we call AIEd for other things, like Semantic Auditing and Information Density to make Adaptive Assessment. It also uses Blooms Revised Taxonomy and Large Language Models to help with Instructional Design and Learning Management Systems