Compare source versions with a change register
A version comparison should identify the two source objects, show the relevant differences, and explain their effect on your research. A redline is a candidate finder. It does not decide whether a change is substantive, whether the documents are comparable, or which note needs updating.
W3C provenance resources provide background on derivation and version relationships. Library of Congress format resources help frame the need to preserve usable source formats. The change register below is a suggested manual workflow, not a claim that a text comparison proves document authenticity.
Pair versions before computing differences
Record title, issuer, period, language, publication date, and document role for both objects. Confirm that they are versions of the same source rather than different report types. A summary announcement and a full annual report can be related without being a suitable line-by-line version pair.
Preserve originals. If you extract text or create a comparison copy, label it as derived working material and link it back to each source. Keep extraction choices visible where they affect the result. Page headers, line breaks, and OCR substitutions can create apparent changes that are not changes in meaning.
Hypothetical change register
Suppose fictional Copper Lake publishes a corrected project notice. The table is an invented example.
| Difference | Classification | Meaning check | Research action |
|---|---|---|---|
| Line wrapping | Formatting | No wording change | Ignore for claims |
| Date revised | Factual field | Read correction explanation | Update timeline dependency |
| “Subject to approval” added | Scope qualification | Read full paragraph | Narrow current summary |
| Section moved | Relocation | Check text retained | Repair locator |
| OCR changes a word | Extraction artifact | Inspect page image | Fix working text only |
The register should preserve the distinction between source change and extraction error. Correcting an OCR copy does not create a new issuer version. Conversely, a real qualification added by the issuer should not disappear just because the overall document looks similar.
Map changes to dependent objects
List the notes, event rows, open questions, and review dates that use the changed field. Update each relevant object and record a short reason. If no active note depends on a change, the register can say so; not every observed difference requires rewriting the archive.
Retain historical notes when their purpose is to show an earlier information set. Add a later-update pointer rather than silently changing their apparent basis. Current summaries and historical records should have explicit roles so a reader knows which one answers the current question.
Verify meaning on the original page
For every change that affects a conclusion, inspect the actual source page in both versions. Read surrounding qualifiers and referenced definitions. A redline can omit the context that makes a revised phrase narrower or broader. Save locators and a concise explanation rather than relying solely on colored highlights.
The comparison is ready when a reviewer can reproduce the pair, identify the difference, and follow its effect into dependent research objects. If the source relationship remains uncertain, report the comparison as provisional. Do not let a clean-looking diff conceal an unresolved identity or period mismatch.
Boundaries and common mistakes
This process is not a legal assessment, a guarantee that all differences were detected, or an investment interpretation. A scoped comparison can be useful if its scope is visible. “Compared the project section” is more accurate than “reviewed all changes” when the rest was not inspected.
Common mistakes include comparing different languages as though they were revisions, overwriting originals, ignoring moved paragraphs, and letting extraction noise dominate the review. Another is changing notes without recording why. A small change register creates a more reliable trail than an unexplained latest version.
A final record check
When a comparison produces many differences, triage them by whether an active claim depends on the affected passage. Inspect those first, then expand if surrounding changes alter interpretation. Keep the triage reason visible. This manages review scope without implying that uninspected differences are harmless or that a small edit is automatically less important than a long addition.
FAQ
Must I compare every page?
No. Define the scope required by the question and state it. Expand when a relevant cross-reference or correction makes additional review necessary.
Is an identical text extraction proof that files are identical?
No. Extraction can omit images, layout, metadata, or other content. Use the comparison appropriate to the question and preserve the original files.