Extracting event graphs from 4M court decisions: enum explosion and nested legal logic
SignificantZebra5883 · reddit · 2026-10-17
A developer building a legal search/analytics engine over 4M Slovak court decisions details why flat representations fail when combining near-100% precision on complex queries with jurimetric statistics. LLM-based structured JSON extraction failed in production for two reasons: (1) enum explosion on open-world facts — LLMs invent thousands of incompatible strings unless locked to enums, which then break prompt adherence; (2) recursive legal logic — nested state chains like CANCELS(REJECTS(REQUESTS(...))) get misread, and flat triples lose evidentiality (alleged vs denied vs unproven). The author proposes an ID-linked event graph (Neo-Davidsonian/RDF-star style) with 8–15 flat event rows per decision, @id references and epistemic qualifiers, with concrete JSON examples, and asks the NLP/IR/LegalTech community for alternatives.
More from coding & agent
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23
- How do you audit AI agent decisions in regulated workflows? A field survey — Ok_Way_1993 · 2026-10-23