Automatic Knowledge Acquisition for Creating Nuclear Juridical Lexica

Abstract Knowledge acquisition constitutes the bottleneck for the creation of legal expert systems. Our prototype CONCAT deals with the problem that the information is formalised to a certain degree by use of legal language. This task is performed in that the process of creating a selective thesaurus for a nuclear juridical lexicon is supported which can be used for automatic indexing and document classification. This selectivity is obtained by distinguishing between precise legal terms and words with fuzzy meanings. The resulting thesaurus can thus represent automatically the expert knowledge of a lawyer about legal terminology. Therefore, CONCAT can be used as a tool for the semi-automatic creation of a legal knowledge base for the automatic representation of the structure and the contents of the documents. As structuring techniques we applied cluster analysis and neural networks. The cluster algorithm creates satisfying results but turned out to be very sensitive as concerns the fine tuning of parameters. The neural network achieves comparable quality without the need for any adaptation.