This paper presents a novel approach for addressing the challenge of blog recommendation in the Web 3.0 era, where traditional frameworks are lacking. The proposed strategy focuses on semantics-driven blog recommendations, achieved by progressively enhancing both query words and dataset terms through knowledge aggregation. Leveraging repositories like NELL and CYC, this approach boosts the richness of preprocessed query words. By integrating knowledge from sources such as Wikidata, DBpedia, NELL, and CYC, the extracted dataset keywords and categories are enriched. The process involves dynamic knowledge stack generation, enabling exponential aggregation of auxiliary knowledge from webpages, online documents, and e-books. This amassed knowledge is efficiently classified using transformer-based classifiers, providing a robust deep learning foundation through a dedicated model. Throughout the model stages, techniques like SOCPMI, KL Divergence, Bose-Einstein index, and Horns index are applied for semantics-driven relevance assessment, facilitating semantic reasoning. Additionally, the Mayfly algorithm is employed to compute optimal instances during intermediate stages, facilitating the alignment between preprocessed query words and enriched dataset entities. An overall precision of 95.42% with an F-measure of 96.17% and an FDR of 0.05 has been achieved by the proposed framework which is a best-in-class blog recommendation framework compared to all other baseline models.