Scaling Out to Become Doctrinal

• Published in: Algorithmic Aspects of Cloud Computing Vol. 10230; pp. 157 - 168
• Main Authors: Panagis, Yannis, Sakkopoulos, Evangelos
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2017; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Big Data; European Court of Human Rights; European Court of Justice; Latent Dirichlet Allocation; MapReduce
• ISBN: 9783319570440; 3319570447
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-57045-7_10

International courts are often prolific and produce a huge amount of decisions per year which makes it extremely difficult both for researchers and practitioners to follow. It would be thus convenient for the legal researchers to be given the ability to get an idea of the topics that are dealt with in the judgments produced by the courts, without having to read through the judgments. This is exactly a use case for topic modeling, however, the volume of data is such that calls for an out-of-core solution. In this paper we are experimenting in this direction by using the data from two major, large international courts. We thus, experiment with topic modeling in Big Data architectures backed by a MapReduce framework. We demonstrate both the feasibility of our approach and the accuracy of the produced topic models that manage to outline very well the development of the subject matters of the courts under study.