Back to top

Semantic Analysis of Court Rulings

Last modified Mar 8


Nowadays, the growing amount of released legislation arises new challenges. The legal but also social relevance of our judicial system presents companies and private individuals the challenge of meeting legal requirements and acting according to the law. The legal conformity of an act is that condition which does not conflict with the current laws. Any interaction of a person or a company with its environment may potentially be regulated by legal norms and thus be subject to jurisdiction. This leads to an increasing number of reached judgments by the various courts.

Legal reasoning constitutes methods that lawyers use to apply laws to facts in a given case in order to answer legal questions. When the meaning of a legal rule is ambiguous, lawyers use legal reasoning to argue for the interpretation that they find most convincing or that is most favorable to their client. The first step in effective legal reasoning is the ability to read a legal rule - such as an article in a code - and figure out how it works. Laws are often written in a way that makes them difficult to understand. Interpretation of the text of law requires a systematic approach. One important factor in this legal reasoning process is the interpretation of these rules in former cases. This is why legal judgments are a crucial artifact within the legal reasoning process.

The digitization provides a continuously increasing number of digitally available judgments. While in Germany, upon request, every judgment must be publicly accessible, only a handful of legal publishers own the majority of digitally available documents. These documents are made available in online databases of the legal publishers. Hereby, a subscription is necessary in order to access the judgments. In addition, other organizations are trying to build up a body of jurisprudence and make it available to a society free of charge.

The increasing digitization of the world and the continuously improving methods and technologies of computer science make it possible to actively support existing processes of legal practice in the identification and interpretation of relevant case law. Nonetheless, information retrieval is not much advanced yet in legal online databases with respect to legal judgments. This results in exhaustive legal research activities for legal practitioners. In the context of this research project the possibilities of both research directions, namely the machine-readable representation of judgements and their enrichment with semantic information, are scientifically examined. Amongother things, it will also be necessary to identify a suitable machine-readable format of judicial texts that influence the information retrieval positively.


Focus and Goals

The Verlyze project aims at the creation of a platform to provide court rulings in a machine-readable manner. Therefore, the first goal is to deduce a suitable format such as an XML schema. The platform needs to support common formats like PDF or .docx as input and hence requires the capability to automatically process the court rulings into the defined format. Hereby, current state-of-the-art natural language processing approaches shall be used to extract semantic information of the underlying rulings. The resulting machine-readable information can be used to provide advanced and more adequate information retrieval to legal practitioners but also act as a database for further research activities aiming at legal reasoning. As the GDPR creates high requirements with respect to data protection, these verdicts also need to be anonymized. As of now, this is also a tedious procedure involving manual labour. As a result, the project also aims at supporting the anonymization of legal rulings. Hence, the project makes research contributions in the following areas:

  1. Derivation of a Machine-readable representation for German legal judgments
  2. Automatic processing of German legal judgments
  3. Extracting semantic information from German legal judgments
  4. Anonymization of German legal judgments


Research Questions

The main research questions within this project address:

  • Which standards exist for representing legal documents and how can these be adapted to define a machine-readable format for German legal judgments?
  • Which NLP approaches are necessary to automatically transform the plain text judgments into this format?
  • How to concept an NLP pipeline in order to extract semantic structures from legal judgments?
  • How to provide computer-enabled support for anonymizing legal judgments?
  • What is a proper software architecture for such a judgment analysis platform, being able to automatically process verdicts?
  • Is a legal court ruling platform beneficial for legal practitioners and researchers?

Contributions (in reverse chronological order)

[Gl19a] Glaser, I.; Landthaler, J.; Matthes,F.: Supporting the Legal Reasoning Process by Classification of Judgments Applying Active Machine Learning, IRIS: Internationales Rechtsinformatik Symposium, Salzburg, Austria, 2019