PLKD: Personalized Learning Content Generation Algorithm Based on Knowledge Granularity Detector
Year: 2026
Type: preprint
Abstract: Personalized learning content generation aims to dynamically construct a sequence of knowledge points and generate corresponding learning materials according to learners’ personal information, goals, and preferences, ensuring alignment with their cognitive level. Existing methods often overlook the rationality of knowledge granularity, resulting in content that is either too abstract or redundant, and making it difficult to ensure clear boundaries between knowledge points. To address these issues, we propose PLKD, Personalized Learning Content Generation Algorithm Based on Knowledge Granularity Detector. First, we adaptively determine the Granularity Threshold using user information and train a Knowledge Granularity Detector on a self-constructed granularity dataset to map any knowledge point to a granularity value within the [0,1] interval. For knowledge points that exceed the threshold and are not specific enough, we perform the Knowledge Points Split and the Peer Boundary Validation to ensure that the splited knowledge points have clear boundaries and reasonable order. Second, in the knowledge point document generation stage, we adopt the Prerequisite Knowledge Points Window mechanism to maintain contextual coherence. Meanwhile, we design the Rare Constraint for the first knowledge point, the Redundancy Constraint and the Hard Prerequisite Constraint for subsequent knowledge points. Through multiple rounds of detection and rewriting, we ensure that the generated content is accurate, non-redundant, and consistent with prerequisite logic. Experimental results demonstrate that PLKD can effectively adapt to users’ personalized needs and provides a new approach for adaptive content generation in intelligent education.
Source: Preprints.org
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7165759405