Legal text classification in Korean sexual offense cases: from traditional machine learning to large language models with XAI insights

Name: Legal text classification in Korean sexual offense cases: from traditional machine learning to large language models with XAI insights
Uploaded: 2026-10-04T12:29:43.093031+00:00
Duration: 376 s
Description: The study compared various AI models for classifying Korean sexual offense legal texts. It found that fine-tuning small language models like KLUE-BERT performed better than large general models, showing domain adaptation is more important than model size. Explainable AI revealed that models often rely on explicit information rather than subtle context to make correct classifications.

The study compared various AI models for classifying Korean sexual offense legal texts. It found that fine-tuning small language models like KLUE-BERT performed better than large general models, showing domain adaptation is more important than model size. Explainable AI revealed that models often rely on explicit information rather than subtle context to make correct classifications.

Open the full episode