Using clustering techniques to identify arguments in legal documents

Abstract

A proposal to automatically identify arguments in legal doc- uments is presented. In this approach, cluster algorithms are applied to argumentative sentences in order to identify arguments. One potential problem with this process is that an argumentative sentence belonging to one specific argument can also simultaneously be part of another, distinct argument. To address this issue, a Fuzzy c-means (FCM) clustering al- gorithm was used and the proposed approach was evaluated with a set of case-law decisions from the European Court of Human Rights (ECHR). An extensive evaluation of the most relevant and discriminant features to this task was performed and the obtained results are presented. In the context of this work two additional algorithms were developed: 1) the “Distribution of Sentence to the Cluster Algorithm” (DSCA) was de- veloped to transfer fuzzy membership values (between 0 and 1) generated by the FCM to a set of clusters; 2) the “Appropriate Cluster Identifi- cation Algorithm” (ACIA) to evaluate the proposed clusters against the gold-standard clusters defined by human experts. The overall results are quite promising and may be the basis for further research work and extensions.