alirezaaminzadeh's picture
Add PermitGuard synthetic bilingual PTW corpus + SIMOPS pairs + eval
aa80965 verified
|
Raw History Blame Contribute Delete
2.43 kB
metadata
license: mit
tags:
  - oil-gas
  - petrochemical
  - hse
  - permit-to-work
  - control-of-work
  - simops
  - risk-classification
  - text-classification
  - persian
  - bilingual
  - synthetic
  - aria-ai
pretty_name: PermitGuard PTW Samples (Synthetic Bilingual)
size_categories:
  - n<1K

PermitGuard — Synthetic Bilingual Permit-to-Work Samples

Part of the Aria AI oil, gas & petrochemical technical-validation portfolio (Aria SafeOps → Control of Work / PTW). Companion model: alirezaaminzadeh/permitguard-risk-classifier and Space: alirezaaminzadeh/permitguard-ptw-risk-classifier.

Data honesty

This corpus is 100% synthetic. There are no real permits, incidents, PII, contractors, or named facilities. Equipment tags such as T-402 / V-101 / P-205B are fictional placeholders reused from the sibling PetroSafe RAG demo unit for portfolio consistency. About 19.2% of permits deliberately mismatch free text vs structured flags (the text claims a precaution that the form did not actually complete, or vice versa) — this is a designed documentation-gap feature, not a generator bug.

Files

File Rows Description
permits.parquet 620 Structured PTW fields + bilingual free text + rule-derived risk_level / fired_rules
simops_pairs.parquet 260 Balanced conflict / no-conflict pairs (5 SIMOPS rules), with JSON of both permit records
eval_results.json — Held-out text-classifier metrics + SIMOPS distribution
benchmark_report.json — Headline KPIs + methodology note
manifest.json — Split IDs, work types, risk levels

Split: 495 train / 125 test, stratified by risk_level. Non-compliant (at least one risk rule fired): 34.2%.

Related

MIT · Aria AI Engineering Team