OmniPage's suspect-word markers help direct proofreading; they do not certify that unmarked text is correct. Use them as one part of a review process that also checks important names, quantities and meaning against the source image. A plausible dictionary word can still be the wrong word.
This guide follows the official OmniPage Ultimate 19.2 help. Keep the original image and a saved working project so corrections remain traceable. Define the level of review the document needs before deciding that clearing every visible marker completes the task.
Understand what the markers represent
The proofreading guide distinguishes suspect characters and words from non-dictionary terms. A specialized name may be flagged despite being correct, while an ordinary-looking word may escape a flag despite changing the meaning of a sentence.
Do not accept or replace every marked term automatically. Compare it with the source and the document's terminology. A quick sequence of confirmations can remove useful prompts without improving the text if the reviewer is judging only whether a suggestion looks familiar.
Review terminology in context
For names, codes and technical terms, use an authorized reference where one exists. A dictionary suggestion is not necessarily the term intended by the source author. Keep uncertain passages visible for the person responsible for the document.
Check important unmarked values
Dates, amounts, reference numbers and negations deserve a separate pass when they affect decisions. These values can be wrong without producing an obvious spelling error. The absence of highlighting is not an independent verification of their accuracy.
Inspect the proofing configuration
The Proofing options reference describes controls for marking non-dictionary words, automatic proofreading and related behavior. Check the active options before interpreting the number of visible flags. Two jobs with different settings can present different marker counts without one having more accurate recognition.
| Review item | Useful evidence |
|---|---|
| Suspect character | Comparison with the source image |
| Non-dictionary term | Approved spelling and document context |
| Important number | Digit-by-digit source comparison |
| Uncertain sentence | Meaning checked against the source, not just spelling |
Work through a short representative page
Use a sample containing ordinary text and the kinds of terms that matter in the collection. Inspect each proposed correction and compare the relevant image area. Record examples of errors that the markers did not catch as well as those they highlighted.
This gives the team a practical review checklist without inventing a recognition accuracy percentage. A small sample can reveal failure types; it does not establish a statistically reliable error rate for an entire archive.
Keep automatic learning under review
The documentation describes training-related behavior associated with proofing. Understand the available settings and their scope before allowing corrections to influence later processing. An incorrect correction can be an unhelpful input to a reused recognition workflow.
For a shared process, assign responsibility for terminology and any supporting resources. Do not let several operators accumulate contradictory local assumptions without a review mechanism. Specialized dictionaries and training resources should serve an identified document set rather than being added merely because they are available.
Distinguish proofing from rewriting
Correct recognition against the source. If the source itself is poorly written or contains an apparent mistake, do not silently rewrite it while calling the result a transcript. Mark the issue and follow the document owner's editorial procedure.
This distinction matters for quotations, records and technical material where a smooth sentence can hide a changed meaning. Keep any authorized editorial revision separate from the recognition correction so later users can understand what changed and why.
Check the saved output separately
After proofreading, save the working project and export the intended output. Reopen that output and inspect critical passages and document structure. Corrections made in the editor should be verified in the actual file the recipient will use.
If a page is omitted, a table is misordered or a paragraph flows incorrectly, clearing word markers will not repair those structural problems. Include page count, reading order and table relationships in the broader acceptance checklist.
Record uncertainty instead of hiding it
Keep a short exception list for passages that cannot be read confidently from the available source. Ask for a better original or a knowledgeable review when needed. Do not replace uncertainty with an invented value merely to finish the job.
Our PDF workflow guide discusses output usability more broadly. For OmniPage proofreading, completion should mean that the agreed review has been performed, important content has been checked and unresolved uncertainty remains explicit—not simply that the page no longer contains colored markers.
For structural errors that word flags cannot fix, see values assigned to the wrong table columns.
