Inteligência computacional aplicada à análise textual para detecção de litigância predatória no Tribunal de Justiça do Tocantins

Abstract

This study presents the development of the Tanatose tool, designed to support the automated detection of predatory litigation within the eProc electronic judicial system, used by the Court of Justice of the State of Tocantins. The increasing judicialization and the ease provided by electronic proceedings have enabled abusive practices, such as the mass filing of lawsuits with substantially identical initial petitions. In this context, a solution based on techniques of Artificial Intelligence, Data Mining, and Natural Language Processing was proposed, focusing on textual comparison between judicial documents. The Tanatose tool automates the retrieval of initial petitions via the National Interoperability Model (MNI), extracts and processes the texts, applies textual similarity algorithms, and generates reports to support procedural audits. Experimental validation, conducted on a set of 304 real cases, demonstrated a significant reduction in screening time, increased accuracy in identifying repetitive patterns, and improvement in the quality of audited information. The research highlights the feasibility of applying artificial intelligence techniques in the legal environment, while respecting the principles of transparency, traceability, and data security. Additionally, it suggests future enhancements such as the historical storage of audits and the adoption of advanced machine learning models, expanding the potential for combating predatory litigation within the Judiciary.