Desenvolvimento de uma ferramenta para segmentação de petições iniciais

Abstract

The growing volume of lawsuits has created challenges in organizing and analyzing documents, making it essential to adopt automated solutions. In this context, this paper proposes the development of a tool for the automatic segmentation of initial petitions, automatically identifying and extracting their three main sections: Fact, Thesis and Request (FTP). The aim is to improve the sorting of these documents, reducing the time needed for analysis and contributing to the decongestion of cases. The tool was implemented using regular expressions (RegEx) to identify textual patterns and automatically segment petitions. To assess the quality of the extraction, two validation methods were applied: Jaccard Similarity and the metrics Accuracy, Recall and F1-score, comparing the automatically extracted passages with the manual annotations made by experts. The results show that automatic segmentation contributes to the automation of legal document sorting, minimizing the need for manual intervention and speeding up textual analysis. Despite the challenges related to the structural variation of petitions, the tool represents an advance in the modernization of the judicial system