In the era of big data, short video has become the main carrier for college students to carry out information exchange and interaction. In this paper, Web3D virtual technology is used to design and realize the Web3D short video learning and education platform. It creates 3D studio models to realize human-computer interaction. The hybrid short video recommendation model of deep learning is combined with the education platform to extract useful data from the massive likes and comments data and make personalized short video recommendations. Simulation experiments were conducted for the learning and education platform and recommendation model, and the study shows that the performance of the short video learning and education platform achieved the expected effect. The deep learning-based short video course recommendation function is perfect and error-free. Compared to other algorithms, the precision rate (60.845), recall rate (56.943), and F1 value (59.349) of the recommendation model in this paper can achieve better results. The interaction method used in this paper’s platform (mean value of 5.2) can attract users more than the interaction method used in traditional teaching (4.8).