Knowledge graphs are essential for many AI applications, but building them is complex and time-consuming. Imagine a world where AI could automatically extract knowledge from unstructured text, creating these valuable knowledge graphs with minimal human intervention. New research explores how Large Language Models (LLMs) can achieve this by using Wikidata, a massive, openly editable knowledge base, as a guiding framework. The challenge with current LLMs in knowledge graph construction is maintaining consistency and connecting generated knowledge with existing databases. This new approach uses Competency Questions (CQs) to identify the core knowledge domain within a text. Essentially, the LLM generates questions about the text and then extracts the relationships and concepts within those questions. These extracted relationships are then compared to existing Wikidata properties. If a match is found, the existing Wikidata property is used; otherwise, a new property is created, ensuring consistency and interoperability with the existing knowledge base. This clever use of Wikidata doesn't just help organize the information. It also taps into the implicit knowledge embedded within LLMs during their training, making the generated knowledge graphs richer and more nuanced. The process is grounded by generating an OWL ontology—a formal way of representing knowledge—based on the extracted relations. This ensures the knowledge graph is structured and machine-readable. Experiments on benchmark datasets show this approach outperforms traditional methods and offers greater interpretability. The research demonstrated an improvement on the SciERC dataset, which includes relations not found in Wikidata. While the performance slightly dipped when the system wasn’t limited to a predefined target schema, this actually highlights its potential to discover new knowledge structures not explicitly defined in existing resources. The future of this research is particularly exciting. Imagine an AI system that can not only answer questions but also provide a clear, auditable trail of how it arrived at that answer. This ontology-grounded approach paves the way for more transparent and interpretable AI, opening new possibilities for applications like robust question-answering systems and integration with vast knowledge bases like Wikidata.