A learning technique for legal document analysis

More and more law is available freely on the Internet. The growing complexity of legal rules and the necessary adaptation to user needs requires better instruments than manual browsing and searching interfaces of the past. Information reconnaissance of an unknown text corpus would provide a major help. Our research on neural networks concerns adaptive learning techniques for information reconnaissance in legal document archives. Self-organising maps offer besides successful classification a promising tool for this purpose. The neural processing elements can be labeled with the most appropriate keywords to describe the contents of the documents. Applying the tools of refinement, our novel approach describes the most interesting features of the document. The user can choose properly between the various units in order to refine the next step of research. An integration of this tool of information reconnaissance into an intelligent agent is straightforward and will bring much benefit in a practical application.