AI that finds the facts hidden in court verdicts

Criminal justice data often record charges and outcomes—but not what actually happened. Yet many continental European verdicts spell out the facts. This study shows we can automatically pull those fact sections from public Slovak court decisions.

The team tested two tools: 1) smart pattern‑matching that recognizes how courts mark the “facts” segment (even when letters are spaced out), and 2) a large language model prompted to extract the description.

Results: a basic method found facts in 40.5% of cases. Advanced patterns reached 97%; the LLM 98.75%; together 99.5%. In checks by law students, the advanced methods matched human annotations about 90–92% of the time (vs 34.5% for the baseline).

Why it matters: richer, standardized descriptions of offenses could power better research, monitoring, and transparency—without waiting for new administrative data.

Paper: http://arxiv.org/abs/2511.05320v1

Register: https://www.AiFeta.com

AI LLM LegalTech Justice Courts NLP OpenData Slovakia Research