Tipologi Argumentasi Jaksa dalam Tuntutan Perkara Kekerasan Seksual: Analisis Corpus Hukum Berbasis Natural Language Processing

Abstract

This study examines the typology of prosecutorial argumentation in sexual violence cases by integrating legal analysis with  Natural Language Processing  (NLP) to identify reasoning patterns embedded in prosecutorial indictments. Existing studies in computational legal research predominantly focus on judicial decisions, while empirical analysis of prosecutorial reasoning remains limited. Addressing this gap, the study proposes a computational approach to analyze prosecutorial arguments as a legal corpus. A qualitative empirical legal design was employed using prosecutorial indictments as the primary data source. The corpus was analyzed through text preprocessing,  Term Frequency–Inverse Document Frequency  (TF–IDF), topic modeling, clustering, and argument mining to identify linguistic patterns and argument structures. The findings reveal four dominant typologies of prosecutorial argumentation: legalistic argumentation, evidence-centered argumentation, victim-oriented argumentation, and integrative argumentation. These typologies demonstrate that prosecutorial reasoning extends beyond formal legal compliance by incorporating evidentiary strength and victim protection into a coherent argumentative framework. The study contributes to the development of computational legal studies by introducing a   computational typology of prosecutorial reasoning  , which integrates legal argumentation theory with NLP-based corpus analysis. The findings also provide practical implications for improving prosecutorial quality assurance, developing AI-assisted prosecution analytics, and supporting transparent and evidence-based prosecutorial decision-making through explainable artificial intelligence.