.png)
Legislation and regulations are quite complicated for both citizens and government professionals. At the same time, we see that AI technology is developing rapidly, but does not simply provide correct answers to legal questions.
In this innovation project, together with a number of government organizations, we are investigating how we can achieve a “WetGPT”. An AI solution that helps citizens, entrepreneurs and government professionals with questions about laws and regulations, tailored to their own specific situation. The starting point is that the answers must be demonstrably correct, explainable and traceable.
Het resultaat
The current GPT models are quite good, but that is not good enough for this purpose. They continue to inherently hallucinate, so a different architecture is needed. One that focuses on knowledge modelling, not GPT.
In this project, we will first create various models of the Woo, Archives Act and AVG information laws. We record these in knowledge graphs (a form of symbolic AI). This creates a legal foundation, which can then be used by various AI techniques (including GraphRag). The research hypothesis is that this architecture will lead to more precision, better explainability and traceability.
We also use the Law Analysis method and make maximum use of existing legal meta-models, standards, previous projects, taxonomies, interreferences, etc.
All knowledge and insights gained will be made public and the results will be adopted by Logius/Koop, RviHH and the Better Working Together Program (BSW).