9:00 AM - 9:20 AM
[2G1-GS-11-01] Generative AI-Based System for Extracting Crime-Constituting-Facts in Criminal Cases
Keywords:GPT, case law analysis, criminal cases, sentence extraction
In criminal law studies, analyzing numerous criminal cases is essential but time-consuming and costly. In our research, we have created a system that automates the analysis using generative AI, specifically utilizing the GPT-4. This system is designed to automatically extract facts from criminal cases that fulfill the requirements for establishing a crime, which are called the elements of crime.
To evaluate the performance of our system, we conducted experiments with four types of crimes. As a result, for most elements of the crime, more than 90 percent of the extractions were at least overlapping with the correct answer. Furthermore, the quality of that had no major problems from the perspective of legal experts.Even when extracting facts differed from the correct parts, several extractions were not entirely incorrect in terms of quality. On the other hand, the cases where the extracted parts are summarized or insufficient extraction need improvement.
To evaluate the performance of our system, we conducted experiments with four types of crimes. As a result, for most elements of the crime, more than 90 percent of the extractions were at least overlapping with the correct answer. Furthermore, the quality of that had no major problems from the perspective of legal experts.Even when extracting facts differed from the correct parts, several extractions were not entirely incorrect in terms of quality. On the other hand, the cases where the extracted parts are summarized or insufficient extraction need improvement.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.