Abstract
In preschool education, integrating children's literature into digital games remains challenging due to difficulties in preserving narrative depth and providing personalized learning. This study designed a game-based learning platform based on semantic analysis and dynamic hierarchical generation, comprising a text parsing engine, a dynamic level generation module, and a personalized learning adapter. Sixty kindergarten children, aged four to six years, were evaluated, with 30 using the platform and 30 using a baseline system. Results showed a 40% reduction in frustration rate (χ2(1) = 4.92, p = 0.027), a learning gain rate of 27% versus 20% for the baseline, and recommendation accuracy of 90% versus 75%. Exploratory comparisons also suggested improved story comprehension, teacher-perceived relevance, narrative-task alignment, and engagement. These findings suggest that the platform can effectively integrate children's literature with game mechanics, offering a feasible digital education solution for early literacy development in preschool settings.