Record report-level conditions beside extracted observations

Context has a scope

Carry the relevant context into the register so reviewers do not have to reopen the report to interpret each row.

Our public material-register example shows a synthetic report beside downloadable output. Use it to inspect how values and source references travel together. The self-service ParseMill trial is for invoice PDFs; supplier test reports and engineering review need a separate fit discussion.

Keep a distinction between reported conditions and actual compliance with them. Transcribing a stated condition does not verify that the procedure was performed as described. The register should represent the source statement without adding that conclusion.

Use a manual context sheet

These invented conditions demonstrate different scopes. Do not propagate C3 across the register until the source establishes its applicability. If a later clarification resolves it, record the new evidence and the observations affected by the assignment.

W3C: RDF Data Cube Vocabulary supports attributes at different levels of an observation structure. Your worksheet can use that idea without assuming every report-level attribute applies universally.

Capture and assign context deliberately

1. Read the cover, section introductions, table captions and notes for relevant conditions. Record exact text and physical page locations.
2. Define the applicable source range using explicit wording or verified document structure. Keep uncertain boundaries in an exception list.
3. Link observations to context records using stable keys. Preserve the raw condition and any separately normalized value or unit.
4. Check observations at each scope boundary and after page transitions. Confirm that a fill-down or join did not cross into another section.

W3C: Provenance model primer distinguishes evidence entities and derived relationships. Record the source note separately from the activity that assigned its context to observations.

Handle conflicts and absent context

If a table note gives a condition different from the cover page, retain both and identify the apparent conflict. A specific exception may be legitimate, but it needs a source-supported scope. Do not silently choose the value that seems more typical.

A blank condition field should remain unstated or unresolved as appropriate. Borrowing a common laboratory condition from another report changes the register from transcription into unsupported inference. Keep any hypothesis in a review note rather than the accepted field.

Verify contextual usefulness

For the fixture, C1 and C2 attach only to their stated ranges, C4 only to S9, and C3 remains unresolved. The exported package should let a reviewer inspect the condition text and the evidence for its scope.

Bring a report with important conditions outside its table to a scoped fit discussion. Agree how those conditions and their evidence would be retained and reviewed before evaluating a wider pilot. The invoice trial does not establish automatic technical-context assignment.