A hybrid AI approach to criminal offense qualification from narrative case descriptions

Recently, modern technology has significantly influenced numerous fields, particularly due to the latest breakthrough in artificial intelligence, providing novel tools for transforming a wide spectrum of professions, including the legal profession. The legal domain has traditionally relied on documents written in natural language and can benefit from these tools by providing lawyers with the means to support processing legal documents and reasoning regarding legal information. This paper presents three pipelines for decision support in criminal law while applying legal provisions to factual case descriptions. A comparative study is conducted to assess how various approaches to extracting legal information from legal texts and reasoning concerning this information can obtain the same conclusions as those made by lawyers. This work investigates how successfully large language models (LLMs) and legal ontology can be employed to obtain case facts from legal texts and to apply these facts to qualify legal offenses according to the criminal code. The first approach relies on raw legal texts and unstructured factual descriptions, allowing LLMs to perform end-to-end reasoning. The second approach introduces a structured pipeline, in which critical elements in legal provisions guide the extraction of case facts, before applying LLM-based reasoning. The third approach employs a formal ontological model for legal provisions and their elements, with extracted attributes represented as instances of this structure and evaluated using an ontology-based reasoner. The systems are evaluated against a database of case law by comparing inferred legal qualifications with those established in judicial decisions. The results reveal that LLMs can effectively reason regarding legal texts, even with minimal preprocessing. However, the hybrid approach employing LLM-based information extraction and ontology-based reasoning performs slightly better, suggesting that structured knowledge representation offers measurable benefits and reliability in legal reasoning tasks. These findings support the integration of symbolic and language-based methods in developing decision-support tools for legal applications.

Authors

Institutions

Publication Details

Journal
Frontiers in Artificial Intelligence
Published
2026-11-17
DOI
https://doi.org/10.3389/frai.2026.1900507
Primary Topic
Artificial Intelligence in Law
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A hybrid AI approach to criminal offense qualification from narrative case descriptions

Marko Marković, Jelena Matković, Stevan Gostojić, Milica Matijević Gostojić
5 links
Frontiers in Artificial Intelligence
Artificial Intelligence in Law
article

A hybrid AI approach to criminal offense qualification from narrative case descriptions

Marko Marković, Jelena Matković, Stevan Gostojić, Milica Matijević Gostojić
article en

Abstract

Recently, modern technology has significantly influenced numerous fields, particularly due to the latest breakthrough in artificial intelligence, providing novel tools for transforming a wide spectrum of professions, including the legal profession. The legal domain has traditionally relied on documents written in natural language and can benefit from these tools by providing lawyers with the means to support processing legal documents and reasoning regarding legal information. This paper presents three pipelines for decision support in criminal law while applying legal provisions to factual case descriptions. A comparative study is conducted to assess how various approaches to extracting legal information from legal texts and reasoning concerning this information can obtain the same conclusions as those made by lawyers. This work investigates how successfully large language models (LLMs) and legal ontology can be employed to obtain case facts from legal texts and to apply these facts to qualify legal offenses according to the criminal code. The first approach relies on raw legal texts and unstructured factual descriptions, allowing LLMs to perform end-to-end reasoning. The second approach introduces a structured pipeline, in which critical elements in legal provisions guide the extraction of case facts, before applying LLM-based reasoning. The third approach employs a formal ontological model for legal provisions and their elements, with extracted attributes represented as instances of this structure and evaluated using an ontology-based reasoner. The systems are evaluated against a database of case law by comparing inferred legal qualifications with those established in judicial decisions. The results reveal that LLMs can effectively reason regarding legal texts, even with minimal preprocessing. However, the hybrid approach employing LLM-based information extraction and ontology-based reasoning performs slightly better, suggesting that structured knowledge representation offers measurable benefits and reliability in legal reasoning tasks. These findings support the integration of symbolic and language-based methods in developing decision-support tools for legal applications.

Frontiers in Artificial IntelligenceVol. 9
University of Novi Sad (RS)
Peace, justice, and strong institutions
Artificial Intelligence in Law
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon