A formula can look familiar and still be wrong. One lost subscript changes which quantity a symbol names; a minus sign read as a dash can change the operation. When reviewing OCR mathematical formulas from a PDF, compare expressions with the scan rather than with the equation you expect to see.
The goal is to verify meaning and usable rendering separately. A correct expression displayed badly needs a different fix from an incorrectly recognized symbol displayed neatly.
Choose a sample with several kinds of notation
Select a page that includes inline notation, a displayed equation, a numbered expression, and the explanation that follows it. If the document uses matrices, integrals, or multi-line derivations, sample those too. A page containing only ordinary prose says little about mathematical recognition.
Keep the original page visible at a comfortable zoom. If the scan itself is ambiguous, mark the uncertainty. Enlarging an image can help inspection, but it cannot restore strokes that were never captured clearly.
Record the original location using both the viewer page and any printed label. Equations often have their own numbering, which gives you another route back to the source. A formula that moved to a different place in EPUB still needs to be traceable to that location.
Inspect small symbols before accepting the whole expression
Check subscripts and superscripts independently of the main symbol. In the illustrative expression x_i², the index and exponent have different jobs. Flattening them into ordinary adjacent characters can lose that distinction.
Common confusions include:
- Zero and capital O; one, lowercase l, and uppercase I.
- Greek nu and Latin v; rho and p in some typefaces.
- A multiplication sign and the letter x.
- Minus signs, hyphens, decimal points, and punctuation.
- Brackets, braces, and the limits attached to a sum or integral.
Check each occurrence in context. A symbol can legitimately mean different things in separate sections, so a document-wide replacement may introduce fresh mistakes. Do not infer that a variable must be the same merely because two glyphs look similar.
Distinguish inline math from displayed equations
Inline expressions must remain attached to their sentence. Verify the words on both sides, punctuation, and any unit immediately following the formula. A line break should not turn part of an inline expression into a separate heading.
Displayed equations need their line structure checked. Inspect aligned equals signs, cases, matrix rows, fractions, and continuation lines. A converter may recognize all the characters while assembling them in the wrong relationship.
Compare the downloaded representation and the rendered view. Markdown editors vary in their support for mathematical notation; CommonMark alone does not define a complete math-rendering system. If the source text appears correct but the display is not, check the destination editor before changing the mathematics.
For output-format syntax and rendering, see PDF equations in Markdown. This review is about what the scan said in the first place.
Keep equation numbers with the right expressions
Trace a reference such as “substituting into (4)” to the intended equation. The number should not have become part of the expression, and two equations should not have exchanged labels.
Then read the surrounding definition: does “where” still introduce the correct variables? Are assumptions, boundary conditions, and units present? A formula can survive recognition while the sentence limiting its use is dropped at a page break.
For a derivation, review adjacent steps together. An isolated line cannot tell you whether a missing term was intentionally eliminated or accidentally omitted. Compare the original explanation before correcting an apparent inconsistency.
Download, render, and keep a review record
PDF Craft can process scanned PDFs and recognize formulas as part of conversion. Use PDF to Markdown when you want editable content, select Markdown after uploading, and download the completed result. Recognition quality depends on the source; there is no document-independent accuracy guarantee.
Maintain a small table with equation location, disputed symbol, source reading, proposed correction, and verification status. For high-stakes calculations, have a suitably qualified reader check the corrected expression before using it. An OCR result is a reading aid, not mathematical validation.
Finish by reopening the file in the editor or reader you will use. Check a copied expression, its equation number, and the paragraph explaining it. Retain the source PDF alongside your research notes so that later quotations and calculations can be checked again.