Open linked data forms a crucial element at the core of Web 3.0, and a novel approach is proposed in this paper to address this need. The proposed model is designed to automate the process of data generation by extracting topics and categories from the dataset, followed by topic modelling and data synthesis through web crawling across the World Wide Web’s intricate structure. This synthesized data is subjected to classification using robust deep learning LSTM and SVM classifiers. The model incorporates various techniques such as Cosimrank and Belle Ahmed’s index, both integral to Imperialistic Competitive Algorithm, to compute semantic similarity and relevance, enhancing semantic orientation and metaheuristic optimization. Through metadata generation, additional localized knowledge is exponentially integrated, further enriching the model’s depth. The classification of metadata using the LSTM classifier enhances the model’s granularity. By incorporating Jaccard similarity and APMI measure, the model infers semantics-related connections effectively. Furthermore, the inclusion of E-book Glossary information significantly accelerates auxiliary knowledge integration, ultimately refining precision, sensor percentage, F-measure, and FDA of the framework. This innovative open-ended linked data generation paradigm seamlessly integrates deep learning for semantic orientation and reasoning, paving the way for a more advanced Web 3.0.