In the era of Web 3.0, a new strategic learning, reasoning, and inferencing model has been proposed for a reliable web video recommendation framework, with a focus on semantics-oriented approaches. This model employs the knowledge aggregation paradigm, incorporating auxiliary knowledge from various heterogeneous sources. The core strategy of this model is Knowledge Pool Derivation, where knowledge is collected and aggregated from ontologies like WikiData and Nell, which provide valuable knowledge entities. These ontologies serve as valuable resources for enriching the recommendations. To achieve semantic understanding, the model utilizes semantics similarity, which is based on the Hellinger distance and explicit semantic analysis. These techniques help to comprehend the meaning and context of video content, enabling more accurate and relevant recommendations. Feature Selection is employed to enhance the model’s efficiency and effectiveness. The Lance and William index is used to identify and select high-quality features that contribute to the overall performance of the recommendation system. This accelerates the model’s learning process and improves the quality of the final recommendations. The proposed model achieves high percentages of precision, recall, accuracy, and F-measure, specifically 93.88%, 95.77%, 94.425%, 94.4058417%, respectively. Additionally, the model achieves a low value of FDR (False Discovery Rate) at 0.07% is achieved