Die Erkenntnis
If you want to use an AI to create, modify or question data, an ontology is crucial.
But I was surprised by the impact of the ontology’s “readability” for getting good results from my conversations with the AI.
Was bedeutet “Lesbarkeit”?
1. Komplexität der Struktur
Your frequently used data objects should have dedicated Nodes instead of working with combinations of nodes & properties or attributes.
2. Selbsterklärende Namen
The data objects should have self-explaining names. The closer the names are to natural language, the better the translation to a formal database request.
Warum funktioniert das?
My best guess: The AI (ChatGPT) is specifically trained on handling natural language.
So if it is readable for humans, it is readable for AI as well.
Plus: Overwriting the impact of the AI’s training data with your own complex ontology plus rules might be like swimming against the stream.
I am happy to discuss and share my prompts: Contact