AI datasheet extraction

Extract equipment identifiers, limits and units into a reviewable register. A useful result preserves the distinction between design conditions, operating conditions and missing information.

Synthetic evaluation input

SYNTHETIC TRAINING DOCUMENT. NOT EQUIPMENT DESIGN DATA.
DS-001 revision B; table 1
Equipment tag: HX-101
Design temperature: 180 deg C
Operating temperature: 120 deg C
Design pressure: 12.0 bar(g)
Operating pressure: 9.5 bar(g)
Material: not stated

Original Inzonex training example, version 1.0. No customer document or real equipment data. No model has been scored on this page.

Expected fields

FieldReference answer
equipment_tag"HX-101"
revision"B"
design_temperature_c180
operating_temperature_c120
design_pressure_barg12
operating_pressure_barg9.5
materialnull

Check a structured response

The comparison runs in your browser against this answer key. It checks exact fields and values, not the accuracy of an explanation or a model's general ability.

Recommended workflow

Use field extraction with source references. Keep design and operating values in separate columns and require an explicit null for absent fields.

A polished summary is not a datasheet register. The output must retain the exact equipment identifier and the meaning attached to each number. Combining a design temperature with an operating pressure creates a record that never existed in the source.

Start with one document revision and a small field contract. A reviewer should be able to find each value without searching the whole document again. Record the revision with the result; two accurate extractions from different revisions can still conflict.

This packet deliberately omits material. The correct answer is not the material commonly used for this equipment. The omission check measures whether the extraction preserves an unknown rather than filling it with a plausible guess.

Failure checks

  • Do not swap design and operating values.
  • Keep gauge pressure distinct from absolute pressure.
  • Return null for material, not a guessed alloy.

What this exercise does not prove

A text extraction pass does not test OCR on scans, table boundaries in a real PDF, or engineering suitability.

For an actual evaluation, keep a separate held-out set, record tool/model version and settings, and log raw outputs, corrections, elapsed time and actual charges. Do not compare tools tested on different inputs as though they ran the same benchmark.

What is checked in AI datasheet extraction?

Extract equipment identifiers, limits and units into a reviewable register. A useful result preserves the distinction between design conditions, operating conditions and missing information.

What does AI datasheet extraction not prove?

A text extraction pass does not test OCR on scans, table boundaries in a real PDF, or engineering suitability.

Related tasks

Match this workflow to your data requirements

Methodology and reuse

These packets and answer keys are original Inzonex educational material, licensed CC BY 4.0. Attribute Inzonex and link to this task page when reusing the packet. The licence does not cover third-party material linked from this site.

NIST AI 600-1: background on generative AI evaluation and risk. This exercise is not NIST-certified.