Development of a semantic and intelligent knowledge graph for well-being energy

Drinkizz, a wellness brand, provides detailed information about ingredients, nutrients, and health benefits, but this knowledge is locked in a static PDF Handbook, making it difficult for staff to access quickly and for customers to explore effectively. To address this challenge, our project aims to design a semantic knowledge graph that transforms the handbook into an interactive AI system.



We first extracted business concepts and relevant relationships from the PDF Handbook by leveraging ChatGPT, then organized them into a structured ontology using Protégé and deployed it in Neo4j with the Neosemantics (n10s) plugin. By using the Neo4j library and GPT-4 to retrieve relevant data and generate answers, users can ask natural language questions and receive accurate, explainable responses backed by verified information. 

The chatbot interface connects users to the knowledge graph in real time, delivering accurate answers instantly. A Graph-RAG workflow orchestrates this process, mapping user questions to Cypher queries, retrieving facts from Neo4j, and grounding the LLM’s responses in verified data. The system is further validated through real business use cases, ensuring practical value for Drinkizz staff and customers.

This approach boosts transparency, improves efficiency, and builds customer trust by grounding responses in the company’s database, while also offering a scalable model for applying knowledge graphs and explainable AI in wellness and other regulated industries.