← back to dashboard · playground · simulator · authoring
Evidence ingestion
Paste plain text or HTML from a court judgment / order / pleading / contract / email chain. The server runs a regex sweep — extracting dates, citations, monetary amounts, party names (the v pattern), and instruments cited — and emits an events.json-shaped skeleton you can paste into a new spike/trace-N/ directory alongside a Catala rule. Heuristic-only, no LLM, no API credits. The author edits the result before saving.
How to use this page
Turn unstructured legal text into structured facts
A trace in this project is a worked example: a rule + a fact pattern + the calculation. The fact pattern lives in a file called events.json and lists the dated events the rule needs (filing date, contract amount, party names, etc.). Building one by hand from a 50-page judgment is tedious. This page does a regex-only first pass — no LLM, no API credits — and returns a structured skeleton you can clean up by hand.
Three steps
Paste text or HTML from a judgment, order, contract, or email chain into the left panel. Or click "Load DEC 001/2025 sample" for a worked example.
Click "Run extraction". The server returns a summary (counts of dates, citations, amounts, parties found) plus a generated events.json skeleton.
Edit the skeleton. The regex sweep is best-effort — it will miss things and over-match things. Review every event, fix party names, and drop the result into a new spike/trace-N/events.json alongside a Catala rule.
What gets extracted
Dates — anything matching common date formats (15 March 2024, 2024-03-15, 15/03/2024).
Citations — neutral citations like [2024] DIFC CFI 042, ADGM and SICC formats too.
Monetary amounts — USD 1,234,567, AED 50,000, $5m, etc., normalised to numbers.
Parties — the v pattern (Smith v Jones ) plus capitalised entity names.
Instruments cited — references to rules and statutes (RDC 38, Singapore IAA s 31, etc.).
No LLM is used. The output is deterministic given the input. You always edit before saving — the regex pass will not get every party name right.
Input — paste text or HTML
Run extraction
Clear
Load DEC 001/2025 sample
No text yet — paste or load a sample.
Extraction
Generated events.json skeleton
Copy JSON
Download
(extraction output will appear here)