The City of Lohja brings document information into efficient use with Oracle Database AI Vector Search and Property Graph technology.
“With the chatbot developed in this project, we are improving our organization’s internal communication and significantly facilitating and speeding up information retrieval for both supervisors and employees. Rapid access to up-to-date guidelines streamlines daily work, harmonizes practices across units, and reduces the internal workload of support services (e.g., HR support and ICT helpdesk), freeing up working time for other tasks and the continuous development of operations.”
Pasi Perämäki
Director of Information Management and Development
City of Lohja
In collaboration with Miracle, the City of Lohja has developed an advanced document chatbot that revolutionizes how city employees utilize their extensive internal document archives.
With a growing volume of unstructured documents, the city required a solution that goes beyond traditional keyword searches
The intelligent assistant operates entirely within the Oracle AI Database 26ai and enables contextual search and utilization of document data using natural language.
The solution’s user interface is built on the Oracle APEX platform, ensuring a secure, responsive, and user-friendly experience. Thanks to APEX, the development team was able to rapidly prototype and deploy the application without a complex architecture. The back-end logic is implemented using efficient PL/SQL, ensuring high performance and data integrity directly within the database.
At the core of the solution is a Graph RAG (Retrieval-Augmented Generation) approach, which outperforms traditional search methods. The system utilizes large language models (LLM) to analyze documents, extracting key entities and their interrelationships. These relationships are stored in the database and modeled using Oracle Property Graph technology. This allows the chatbot to ”understand” the connections between different municipal projects, decisions, and stakeholders, rather than treating documents as isolated pieces of text.
Simultaneously, the solution utilizes Oracle AI Vector Search functionality for semantic similarity searching. By converting document text into vectors, the database can identify relevant information based on the meaning of the question, even if exact keywords are not present.
By combining advanced graph and vector queries, the chatbot provides contextual information. It not only finds the right documents but also reveals related insights that would remain undetected with traditional RAG solutions. The result is a powerful tool that saves time and supports better decision-making by providing comprehensive and accurate answers utilizing the city’s entire knowledge base.