HILDA: knowledge extraction from neural networks in legal rule based and case based reasoning

A major requirement for legal expert systems involved in generating legal advice and purporting to adjudicate on disputes is that they explain their reasoning. Even systems involved in predicting the outcomes of legal disputes are enhanced by this facility. Difficulties in extracting knowledge from neural networks (“NNs”) have made their application to legal expert systems somewhat limited. HILDA incorporates some aspects of rule based reasoning (“RBR”) and case based reasoning (“CBR”) to assist the user in predicting case outcomes and generating arguments and case decisions. The system can use the NN to guide RBR and CBR in a number of ways. Knowledge extracted from a NN could also be used to iteratively refine the system´s domain theory. This refined domain theory is one way in which HILDA can carry out RBR and CBR