Typical scenario

OCR errors are treated as the writer's habits

If authorship comparison uses OCR versions of scanned letters, repeated letter substitutions and broken words may originate in the same recognition program. In this typical case they entered a linguistic feature table. Check the page images before treating repeated errors as individual habits.

What needs to be clarified

Verify OCR against letter images, distinguish recognition and manual correction changes and assess the remaining linguistic features' suitability for the authorship question.

Materials to prepare

  • All page images at original resolution and in order
  • Recognised text before and after proofreading
  • The report, feature table and recognition information

What to compare

Locate each disputed spelling on its page; check letter legibility, hyphenation, punctuation and line boundaries. Compare recurring software errors across pages and mark manual corrections separately. Use verified transcription for linguistic comparison, retaining ambiguities and excluding guesses for unclear characters.

Possible outcome of the analysis

A checked transcription and error-provenance table may show effects on the earlier opinion; some features may be excluded and authorship remain unresolved.

Limits of the conclusions

The same OCR error does not prove common authorship. An illegible character cannot be confidently inferred from context and then treated as a writer-specific feature.

Authorship examination: text authorship, anonymous messages, correspondence, style and expert report reviews
Write to usDiscuss your task ↗