Back to search

University of the Aegean

Legal Text Mining Using Big Data Processing in High Performance Computing Infrastructures

Abstract

dc:description

This doctoral dissertation presents a comprehensive investigation and development of an advanced legal information system, addressing the growing need for effective and intuitive access to legal information within an increasingly complex regulatory environment. The research mixes theory with real practice and uses new technologies so the platform developed can serve legal study and analysis more effectively. The study begins by identifying key challenges in the field of legal informatics, focusing mainly on the issues of information overload and the difficulty of understanding legal content. The challenges come from two main sources: first, the enormous quantity of legal material now online, and second, the highly technical style of legal language, which non-experts often find difficult to follow or make sense of. The questions raised here touch on design theory and how it can help in building a legal information system for users. They also deal with the use of artificial intelligence, natural language processing, and with the role of open-data principles in keeping information transparent and accessible. As a basis for the research, the work turns to a review of what has been done so far in legal text mining and in legal information systems more broadly. Among the initiatives examined are OPENLAWS, EUCases, and EUCASES. Looking closely at these systems made it possible to see where good practices already exist, but also where gaps remain, insights that shaped both the design and the functionality of the system proposed here. The methodology followed, used several methods at the same time: theory, system design, building the system, and testing it in practice. This combination helped make sure the final system was based on theory but also worked for real users. Because the design was done in repeated cycles, and feedback was collected at every stage, the system could be improved little by little during the study. One important part of the study was to build a single metadata schema for legal information and for that purpose, it brought together elements from well-known ontologies, including the European Legislation Identifier (ELI), Akoma Ntoso (AKN), and the Data Catalog Vocabulary (DCAT). The schema developed here gives a clear way to describe and organize legal documents and this makes searches easier to run and results more accurate. The system was put together in a modular fashion so it can grow when needed. It has two main sides: an offline part that looks after data collection, processing, and storage, and an online part that provides the services and the user interface. The main technologies are the Blazor framework with C# (.NET 6), Elasticsearch for powerful searching, and a set of visualization libraries for interactive displays. A notable feature of the system is its search and filtering: it does more than simple keyword matching, since it takes into account context and the semantic links between legal concepts. This makes it possible for users to run complex searches across frameworks and even across languages. The system includes several visualization tools that improve how legal information can be presented and one example is the timeline view, which helps readers see how laws change over time. Other tools show legal relationships as graphs and allow text comparisons to highlight differences between documents. The study also looks at the problem of multilingual support in legal information systems. To deal with this, services such as the EU's eTranslation API and the IATE terminology database were integrated, giving reliable translations of legal terms and documents in a range of European languages. Part of the dissertation looks closely at the design of the user interface. The goal was to keep the system powerful but still easy to use, so it could work for both professionals and the general public. Features include custom dashboards, teamwork tools, and interactive graphics, which improve the experience and help with legal research. The work also relied on high-performance computing (HPC) when handling and marking very large collections of legal texts. This made processing much faster and showed that the system can scale to deal with big data in the legal domain. The study also looks at ethical questions and at the limits of using advanced technologies in legal information systems. Among the issues considered are data protection, bias in algorithms, and the possible effects on legal practice. The need for careful and responsible development of such systems is underlined. Two working versions of the framework are described: the proposed platform itself and the MarinDoc system. These show that the framework can be adapted to different areas of law and to various user requirements. The study's results are important. The combination of advanced technologies into a single platform that is easy to use is a real step forward for legal informatics. Parallel search across frameworks, visualization functions, and open-data ideas all add to better access and more transparency. The research is not without its limits. Among them are the difficulty of keeping legal databases complete and up to date across many jurisdictions, the limited accuracy of machine translation in handling legal terms, and the fact that the system cannot yet reflect all the subtleties of legal reasoning. For the future, the dissertation points to several directions: a wider database, stronger NLP tools, better machine translation for legal documents, and the use of predictive analytics in a more advanced form. The dissertation contributes to legal informatics by presenting a system that combines strong search, clear visual displays, and an interface designed with the user in mind. The result is that legal texts, often hard to deal with, become easier for different kinds of users to access and understand. This can lead to greater openness and efficiency in legal practice while at the same time the work accepts that limits remaining, but it offers ground for further development in a field that is moving fast. In explaining the system, the dissertation sets out the technologies it relies on. Blazor with C# (.NET 6) is the core framework and was used to build the interactive web platform. For storing data, MongoDB was adopted, and for search, Elasticsearch, so that even very large volumes of legal texts can be handled and retrieved quickly. Natural Language Processing (NLP) tools such as spaCy and BERT were integrated so the system can process and analyze complex legal texts. They are used for keyword extraction, document summaries, and better contextual understanding in searches. For visualization, D3.js was included to create interactive graphics that display legal links and timelines in a clear way, giving users a better grasp of the information. The architecture follows a microservices model and makes use of Docker for containerization. This setup allows the system to be scaled easily and lets different components be developed, updated, and deployed independently. Apache Hadoop is used for distributed storage and processing, which lets the system manage very large sets of legal data as they continue to grow. TensorFlow was chosen for machine learning, so the system can learn from users and improve its search results, making them more accurate and relevant over time. The dissertation also deals with security and data protection. The system makes use of encryption together with secure login methods and to be more precise, OAuth 2.0 is applied to manage API access, while JWT (JSON Web Tokens) keep user sessions safe. For hosting and scaling the system relies on Microsoft Azure. That choice makes it easier to deal with changing workloads and keeps the service reliable. One challenge the dissertation identifies is interoperability with other legal tools and databases. The solution is the use of APIs and data-exchange protocols, making it possible to connect with research platforms, court management systems, and legal practice software. Another strand of the dissertation considers ethics and the bias that may arise when AI is applied to legal information. Here the study suggests creating an ethics board and using strict testing rules so that AI components stay transparent and free of bias in decision making. Another area examined in the study is comparative legal analysis. The system offers tools that allow researchers to compare legal approaches across jurisdictions and to support work on international law and harmonization. Toward the end, the dissertation considers the use of virtual and augmented reality for teaching law and for visualizing legal material highlighting that they can be used in legal education and in showing legal concepts, giving students and researchers an immersive way to understand complex cases and past legal events. To conclude, the dissertation presents a structured approach to the development of legal information systems. It combines current technologies with attention to the specific demands of the legal field, offering ways to improve access, understanding, and analysis of legal materials. The system demonstrates how professionals, researchers, and citizens can engage with legal information in more effective ways, supporting transparency and efficiency. With the continuing evolution of legal informatics, this work serves as a foundation for further progress and for the thoughtful use of technology in law.

Degree

thesis:*
Grantor dc:publisher
University of the Aegean
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Loutsaris, Michalis-Avgerinos
  • Λουτσάρης, Μιχάλης-Αυγερινός

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/59728

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Loutsaris, Michalis-Avgerinos; Λουτσάρης, Μιχάλης-Αυγερινός. Legal Text Mining Using Big Data Processing in High Performance Computing Infrastructures. University of the Aegean, 2024. http://hdl.handle.net/10442/hedi/59728