Text Converter Tools
Evidence-backed workflow test

Cleaning Text Copied from PDF: A 12-Sample Paste Artifact Test

Reproducible fixtures for PDF copy/paste artifacts—hard wraps, soft hyphens, headers, spacing—tested with local transforms. This site does not parse PDF files.

Question
After copying text out of a PDF, which cleanup steps are safe to automate, and which still need a human?
Samples
12 fixtures
Generated via core
8
Manual / limits
4
Published
Modified
Author
Dionis Ceban
Method
Twelve synthetic paste samples (not PDF bytes) were run through the shared JavaScript core used by Trim Text, Remove Extra Spaces, Remove Empty Lines, Remove Duplicate Lines, and Text Cleanup Studio. Observation-only fixtures record limitations without inventing a repair.

Why this test matters

PDF extractors and viewer copy often inject wraps, soft hyphens, running headers, and spacing noise. Cleaning too aggressively can merge real hyphenated words or delete meaningful numbers.

Tools under test

What automation can and cannot do

Safe to automate (in this corpus)

  • Trim line edges
  • Collapse runs of horizontal spaces/tabs/nbsp
  • Collapse excessive blank lines
  • Deduplicate exact repeated header lines after trim

Needs manual review

  • Soft-hyphen (U+00AD) removal vs real hyphenated compounds
  • Deleting digit-only lines that might be page numbers or data
  • Re-flowing hard-wrapped sentences into paragraphs

Not repaired by these tools

  • Multi-column reading order
  • Missing spaces that glued words together
  • Recovering layout, fonts, or tables from the original PDF

Result summary

8 of 12 fixtures were executed through assets/js/tools-src/core.js. 4 entries are labeled manual review or limitation without inventing a transform result.

Full test matrix (12 samples)
IDCaseInputOperationObservedStatus
pdf-hard-wrap-01 Hard line wrapping mid-sentence This paragraph was wrapped at arbitrary column widths so copied text breaks mid sentence. identity This paragraph was wrapped at arbitrary column widths so copied text breaks mid sentence. Pass
pdf-soft-hyphen-02 Soft hyphen at line break This is an exam­ple of a soft­hyphen artifact. Observation (no auto transform) Input contains U+00AD soft hyphens between “exam” and “ple”. Automatic de-hyphenation can destroy real compounds. Manual review
pdf-header-repeat-03 Repeated running header Quarterly Report Body line one. Quarterly Report Body line two. trimText → removeDuplicateLines Quarterly Report Body line one. Body line two. Pass
pdf-page-numbers-04 Standalone page numbers Introduction 12 Next section 13 Observation (no auto transform) Lines that are only digits (12, 13) look like page numbers but could be data. No safe auto-delete without context. Manual review
pdf-excess-spaces-05 Excess horizontal spaces and tabs Word1 Word2 Word3 Word4 removeExtraSpaces Word1 Word2 Word3 Word4 Pass
pdf-blank-lines-06 Excessive blank lines Paragraph A. Paragraph B. removeEmptyLines Paragraph A. Paragraph B. Pass
pdf-real-hyphen-07 Legitimate hyphenated word well-known example state-of-the-art model trimText → removeExtraSpaces well-known example state-of-the-art model Pass
pdf-multicolumn-08 Multi-column copy order Alpha col Beta col A1 B1 A2 B2 Observation (no auto transform) Copied reading order interleaves columns (Left1 Right1 Left2 Right2). No transform on this site can recover layout intent. Limitation
pdf-bullets-09 Bullet characters • First item • Second item · Third item trimText • First item • Second item · Third item Pass
pdf-unicode-punct-10 Unicode punctuation “Smart quotes” and an em—dash with … ellipsis. identity “Smart quotes” and an em—dash with … ellipsis. Pass
pdf-joined-words-11 Words joined incorrectly thequick brownfox jumps Observation (no auto transform) Missing spaces (thequick brown) cannot be repaired reliably without a dictionary and language model. Limitation
pdf-paragraph-bounds-12 Paragraph boundaries after wrap cleanup First paragraph line. Second paragraph line. trimText → removeExtraSpaces → removeEmptyLines First paragraph line. Second paragraph line. Pass

Edge cases and failures

Soft hyphens, digit-only lines, and multi-column interleaving are the failure modes that look “almost fixable” but are not safe defaults.

Conclusion

Treat PDF paste as text hygiene: normalize spacing and blanks first, dedupe only exact repeats, and keep hyphen and page-number decisions manual. TextConverterTools never opens or repairs PDF files—only text you paste.

Limitations

  • Fixtures are synthetic paste strings, not bytes extracted from real PDF files.
  • No OCR, font, or layout recovery is attempted.
  • Language-specific hyphenation rules are out of scope.

Reproducibility

Download fixtures.json and expected-results.json from this page. Re-run with npm run guides:evidence against assets/js/tools-src/core.js. Sample count must remain 12.

Downloadable evidence

Related experiments

All guides