The HOLJ corpus: supporting summarisation of legal texts

Abstract

We describe an XML-encoded corpus of texts in the legal domain which was gathered for an automatic summarisation project. We describe two distinct layers of annotation: manual annotation of the rhetorical status of sentences and an entirely automatic annotation process incorporating a host of individual linguistic processors. The manual rhetorical status annotation has been developed as training and testing material for a summarisation system based on the work of Teufel and Moens, while the automatic layer of annotation encodes linguistic information as features for a machine learning approach to rhetorical status classification. 1 Project Overview