Skip to Main content Skip to Navigation

Méthodes D'Analyse Sémantique De Corpus De Décisions Jurisprudentielles

Abstract : A case law is a corpus of judicial decisions representing the way in which laws are interpreted to resolve a dispute. It is essential for lawyers who analyze it to understand and anticipate the decision-making of judges. Its exhaustive analysis is difficult manually because of its immense volume and the unstructured nature of the documents. The estimation of the judicial risk by individuals is thus impossible because they are also confronted with the complexity of the judicial system and language. The automation of decision analysis enable an exhaustive extraction of relevant knowledge for structuring case law for descriptive and predictive analyses. In order to make the comprehension of a case law exhaustive and more accessible, this thesis deals with the automation of some important tasks for the expert analysis of court decisions. First, we study the application of probabilistic sequence labeling models for the detection of the sections that structure court decisions, legal entities, and legal rules citations. Then, the identification of the demands of the parties is studied. The proposed approach for the recognition of the requested and granted quanta exploits the proximity between sums of money and automatically learned key-phrases. We also show that the meaning of the judges' result is identifiable either from predefined keywords or by a classification of decisions. Finally, for a given category of demands, the situations or factual circumstances in which those demands are made, are discovered by clustering the decisions. For this purpose, a method of learning a similarity distance is proposed and compared with established distances. This thesis discusses the experimental results obtained on manually annotated real data. Finally, the thesis proposes a demonstration of applications to the descriptive analysis of a large corpus of French court decisions.
Document type :
Complete list of metadatas

Cited literature [221 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Thursday, October 15, 2020 - 8:34:09 AM
Last modification on : Saturday, October 17, 2020 - 3:05:11 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02967503, version 1



Gildas Tagny Ngompe. Méthodes D'Analyse Sémantique De Corpus De Décisions Jurisprudentielles. Autre [cs.OH]. IMT - MINES ALES - IMT - Mines Alès Ecole Mines - Télécom, 2020. Français. ⟨NNT : 2020EMAL0002⟩. ⟨tel-02967503⟩



Record views


Files downloads