Xu, Huihui; Savelka, Jaromir; Ashley, Kevin
Abstract: In this paper, we assess the use of several deep learning classification algorithms for the purpose of automatically preparing succinct summaries of legal decisions. Short case summaries that tease out the decision’s argument structure by making explicit its issues, conclusions, and reasons (i.e., argument triples) could make it easier for the lay public and legal professionals to gain an insight into what the case is about. By enabling such access to legal resources and lowering legal costs associated with legal information retrieval we hope to improve access to justice. We have obtained a sizeable dataset of expert-crafted case summaries paired with full texts of the decisions issued by various Canadian courts. As these case summaries are quite long, our goal is to (i) identify the core subset of summary sentences (i.e., the argument triples) to make them more succinct, and (ii) train a system that will be capable of creating the succinct summaries directly from the full-texts by extracting the argument triples sentences. As the manual annotation of the full texts is prohibitively expensive, we explore various ways of leveraging the existing longer summaries which are much easier to annotate. We compare the performance of the systems trained on the annotations that are manually ported to the full texts from the longer summaries to the performance of the same systems trained on annotations that are projected automatically. The results are promising (macro F1 = 0.46) and suggest that annotating the longer summaries instead of the full texts of the decisions is a viable strategy worth pursuing further.