Information extraction from legal texts: the potential of discourse analysis

There is an urgent need to automatically identify information in legal texts. In this paper, we argue that discourse analysis yields valuable knowledge to be incorporated in text processing systems. Knowledge about discourse patterns has already been applied in legal text generation systems. But, it is equally important to incorporate this kind of knowledge in legal information extraction systems. This knowledge is helpful for locating information in texts. Also, we demonstrate the need for adequate, maintainable, and possibly sharable knowledge representations of discourse patterns. The findings are illustrated by explicating the role discourse analysis played when building the SALOMON system, a system that automatically abstracts Belgian criminal cases.