In the digital age, millions of organizations accumulate vast amounts of data in specialized archives. However, extracting valuable information from these repositories remains a major challenge, especially when technical staff lack the skills to write structured queries. This problem is magnified in fields such as neuroscience research, biomedicine, or engineering, where domain semantics are complex and metadata follows specific vocabularies. The solution, according to a recent study, lies in combining formal ontologies with large language models (LLMs). The system called NLKGQ (Natural Language Knowledge Graph Query) shows that it is possible to generate accurate SPARQL queries from natural language questions, without additional training or retrieval augmentation. And it does so thanks to careful ontology design and the power of LLMs.
In the business environment, the ability to answer ad-hoc questions about historical data can make a difference in decision making. For example, a CFO might want to know 'what was the profit margin per product in the last fiscal year?' without relying on an IT team. This is where Business Intelligence solutions like Power BI come into play, which Q2BSTUDIO frequently implements for clients in various sectors. However, Power BI requires predefined queries or prepared tables; the NLKGQ approach allows fully open-ended questions, combining the best of both worlds. By integrating a knowledge graph with an LLM, any user can ask questions in their own language and get immediate answers, without needing to know SPARQL or SQL. This democratizes data access and accelerates analysis cycles.
The NLKGQ approach is relevant not only for academia but for any company that manages large volumes of data with defined semantic structures. At Q2BSTUDIO, as a software development and technology company, we see a clear opportunity to apply these principles in business solutions. Creating custom software that integrates natural language and knowledge graph databases can transform how organizations interact with their data. For instance, a business analyst could ask 'what were last quarter's sales by region?' and get an immediate answer without knowing SPARQL or SQL. This capability is especially useful when combined with BI tools like Power BI, which Q2BSTUDIO implements for its clients on AWS or Azure cloud environments. The cloud infrastructure allows the system to scale on demand, maintaining low response times even with massive data volumes.
The secret to NLKGQ's success lies in two fundamental factors: human-readable entity names and well-defined semantic annotations. According to the study, these elements influence accuracy more than the choice of language model or prompt engineering. This underscores the importance of investing in rigorous ontological modeling. OWL ontologies capture not only vocabulary but also semantic relationships between concepts, allowing the LLM to correctly interpret the question's intent. In contrast, traditional relational databases with SQL DDL offer less semantic context, resulting in inferior performance for automatic query generation. For companies, this means migrating to knowledge graph architectures, with help from experts in artificial intelligence, can be a competitive differentiator.
Furthermore, AI agents can act as intelligent intermediaries: they receive the question, translate it to SPARQL, execute the query, and return the answer in natural language. These agents integrate into virtual assistants, corporate chatbots, or self-service portals. At Q2BSTUDIO we have developed prototypes of AI agents that use local LLMs to ensure privacy, a recurring concern in sectors such as banking or healthcare. Cybersecurity is another pillar: by running models on institutional hardware, data never leaves the organization. Our cybersecurity services audit and protect these infrastructures, ensuring that natural language access to data is as secure as any other corporate system.
The NLKGQ system was tested on a large-scale neuroimaging archive, achieving 100% accuracy on a set of competence and regression questions. The researchers used local LLMs running on modest institutional hardware, addressing privacy concerns for human subject data. This is crucial for sectors like healthcare, where confidentiality is paramount. At Q2BSTUDIO we offer cybersecurity solutions that ensure sensitive data remains protected while harnessing the power of generative AI for advanced queries. The integration of specialized AI agents for question interpretation and query execution is another service we are developing for companies that want to automate their data analysis processes.
From a technical perspective, the NLKGQ system consists of a web interface where users pose natural language questions. A domain-agnostic harness translates those questions to SPARQL via an LLM and executes them against a knowledge graph. The development process begins with capturing vocabulary and semantics in a formal OWL ontology. Then, domain-specific code extracts metadata from archive sources and imports it into the graph. Both the ontology and the code are designed to be reusable across domains, significantly reducing implementation effort in new contexts. This modular architecture is similar to what we employ at Q2BSTUDIO for cloud computing projects, where we deploy microservices on AWS or Azure to ensure scalability and flexibility. Additionally, integration with BI tools like Power BI allows real-time visualization of query results, generating dashboards that update automatically with each new question.
For organizations still undecided between relational databases and knowledge graphs, the study compared SPARQL performance against automatically generated SQL. The results were conclusive: OWL ontology, with its structural features (classes, properties, constraints), provides a substantial advantage over SQL DDL when generating queries via LLMs. This is because explicit ontology semantics reduce ambiguity and help the model understand the meaning of the data. In practice, companies that adopt knowledge graphs—whether for product catalogs, document repositories, or customer data—can offer their employees more democratic access to information without the need for advanced technical training. Q2BSTUDIO helps its clients design these ontologies and implement data extraction and loading processes, ensuring the system works from day one.
The future of data querying lies in combining generative AI and ontologies. NLKGQ is an excellent example of how careful design can achieve perfect results in complex tasks. At Q2BSTUDIO we are convinced that this approach will spread across multiple sectors. That is why we offer consulting and implementation services for natural language query systems, integrating artificial intelligence, cloud, and automation technologies. Our team works with OWL ontologies, knowledge graphs, and LLMs to build solutions that allow any user to ask and get accurate answers from their data, without technical barriers. Process automation, combined with the power of AI agents, ensures that even complex questions are answered in seconds, transforming how companies exploit their information capital.
In conclusion, NLKGQ demonstrates that natural language access to metadata is already a reality, thanks to the synergy between well-designed ontologies and advanced language models. Companies that want to lead in the data era should consider this approach as part of their digital strategy. Whether developing custom applications, improving their BI systems, or implementing AI agents, the key is to combine the semantic richness of ontologies with the power of LLMs. At Q2BSTUDIO we are ready to accompany that process, offering from ontology definition to cloud deployment, always with a focus on data security and privacy. Natural language querying is not the future: it is the present, and it is within reach of any organization that decides to take the step.
