Convert scanned paper to Word, how to keep formulas and tables together

Key points of this article

The scanned document does not have a text layer and must be OCRed before exporting. The verification after recognition affects the result more than the export itself.
Formulas are divided into three categories according to their complexity: simple verification and correction, complex ones retaining the original image or re-entering
The table must be read cell by cell based on the original picture. Misplaced numbers, missing decimal points and negative signs are dangerous errors.
Check numbering and cross-references of formulas and tables to avoid text pointing to the wrong location
The results of machine recognition need to be manually compared with the original text, especially experimental data and formulas, before being used for citation or submission.

You have a scanned paper in your hands, and the text, formulas, and tables are mixed together in the PDF. I want to convert it into Word and continue to revise or take notes. The common results are: the text can be basically recognized, the formulas become garbled or blurry pictures, and the rows and columns of the table are misaligned. How to convert scanned documents to Word and preserve the format is a problem that many people who read documents, revise manuscripts and organize materials will encounter.

This article is written for readers who need to process scanned versions of papers. After reading, you will know how to do OCR in the PDF editor of the OmniDoc desktop client first, then export it to Word, then check the formulas and tables in a fixed order, and determine which areas need to be redone manually.

Each page of the scan is an image, and there is no text that can be copied, so the key to conversion is not the export step, but the verification after recognition. If you do this step solidly, you will be much more likely to keep the formulas and tables.

When converting a scanned paper to Word, how can I keep the formulas and tables together? One sentence conclusion

Let’s talk about the conclusion first: leave the text to OCR for recognition. After exporting the table, compare it with the original image cell by cell. If the formula can be recognized, check and correct it. If the formula cannot be recognized, keep the original image or re-enter it in Word. Don’t expect to submit the manuscript directly after exporting it once.

your situation	Recommended practices	reason
Main text, few formulas	Export to Word after OCR, read through and proofread	Text recognition is relatively stable, and the main risks are typos and line breaks.
Contains simple formulas, such as one line, without nested superscripts and subscripts	After recognition, compare and correct it side by side with the original image.	Simple formulas can usually be restored, so focus on checking symbols and subscripts.
Complex formulas, such as multi-level fractions, matrices, summation symbols	Keep the original image or re-enter it in the UniDoc web editor	Recognition errors are often difficult to detect, and manual rewriting is more reliable than modifying garbled characters.
The table has merged cells or spreads across pages	After exporting, rearrange it in Word and check it box by box based on the original image.	The table structure is easily misaligned, and numerical errors are difficult to find.
Papers to be submitted or cited	Keep the original PDF and every version of the Word file at the same time	When disputes arise, you can return to the original text for comparison, and changes can also be traced back.

First determine what kind of problem it is

After the conversion, first see clearly which category the problem belongs to, and then decide to practice the method.

symptoms	Common causes	How to confirm this
Formulas become garbled or incorrect symbols	OCR mistakenly interprets mathematical symbols as letters or numbers, resulting in lower resolution of the original images	Enlarge the original image and compare the characters in the same position in Word
The formula becomes a blurry picture	The formula is not recognized as editable content and is kept as an image as a whole	In Word, click on formulas to see whether it's text or images
Misaligned rows and columns in the table, incorrect number of columns	Table lines are unclear, and adjacent columns are merged or split	Count the rows and columns in the original table and compare them one by one with the Word table
Numbers with fewer decimals or minus signs	OCR is insensitive to fine symbols	Reading frame by frame, compared with the original image
The paragraph order is disrupted or the line breaks are incorrect	The double-column layout was incorrectly stitched in the order of reading	Check whether paragraphs in Word are arranged in the original order of the left and right columns
Headers, footers, and page numbers are mixed into the main text	The scanned page itself contains these elements	Check for any unnecessary content at the beginning and end of each page

Inspect and fix each item one by one

First, confirm whether the PDF is a scanned copy. Try selecting a paragraph of text in the PDF and copying it. If the selection is not correct or the copy is empty, it means the content is an image and OCR is needed first. If the selection works properly, you can export it directly without needing recognition.
Open the PDF with the OmniDoc desktop client and perform OCR on the scanned pages. When there are many pages, you can first identify the first few pages to check the effect, then process the entire text to avoid reworking the entire article.
First correct obvious typos. Directly modify paragraph text in the PDF editor, focusing on changing quantity names (such as x, y, α) and numbers. Errors in identifying these places greatly affect understanding.
Export to Word. Before exporting, confirm that there are no obvious issues with the page order and column divisions. After exporting, open it in Word and read through it once. Don't rush to adjust the formatting; first make sure the content is complete.
Check the formula one by one. Each formula is compared side by side with the original figure, checking the number of symbols, subscripts, and brackets. Edit it directly if possible; If you can't change it, keep the original image, or re-enter it in the UniDoc web editor (app.unidoc.top). UniDoc supports formula editing and can be converted with Word, making it suitable for organizing complex formulas into editable formats.
Handle the form. Simple tables can be checked frame by frame directly. Merging cells and spreading tables usually need to be manually adjusted in Word. When verifying, use the original image as the standard, read each frame by frame, and do not modify based on impressions.
Check the number and location. Confirm that formula numbers (e.g., formula (3)) and table numbers (e.g., Table 2) are consecutive, and that references in the main text still point to the correct position.
Read the entire text against the original text. Pay special attention to key figures such as experimental results and sample size. The results recognized by the machine need to be manually verified, and only after confirmation are they used for reference or submission.

How to Prevent It from Happening Next Time

Try to use clear images when scanning. If you have the original PDF, prioritize using it; Documents, formulas, and small fonts taken by mobile phones are prone to errors in recognition.
For papers with many formulas, identify and check the formula pages separately first, rather than mixing them with the main text to make it easier to locate errors.
When there are many tables, immediately fix the table structure after exporting to Word before writing the main text to avoid discovering row and column misalignment during later rework.
Keep the original PDF and each Word version, and save a separate copy before modification for easy review.
For important numbers and formulas, create a separate checklist, tick each item, and use it after checking.

❓ Frequently Asked Questions (FAQ)

1. Image recognition formulas in Word, why can only images be obtained?

If the formula structure is complex or the original image is not clear during recognition, the formula may be kept as an image. Click on a formula in Word to see if it can be edited. If it's an image, manually rewrite it as the original image, or keep the image as a backup.

2. How to copy a formula in Word while keeping its original format?

Whether the format can be kept depends on whether the formula is editable in Word. There is no formula for copying the scanned footage; it can only be identified first, then compared with the original image or re-entered. Copying directly from the scanned image usually cannot preserve the original format.

3. How do you keep the format when converting scanned footage to Word? Is it necessary to use OCR first?

If the PDF can select text, it can be exported directly without OCR. The scanned document does not have a text layer, so OCR must be done first. The key to preserving formatting is to identify and check columns, tables, and formulas, rather than just checking whether the export was successful.

4. Can machine-recognized paper content be submitted or cited directly?

It is not recommended to use them directly. Identification results may contain typos, and numbers and formulas may also be incorrect. Before submitting or citeing, check item by item against the original text, especially experimental data and formulas.

📝 Summary

Scanned files are converted to Word to retain format; the key is verification after recognition. First perform OCR, then export, and then compare items item by item in the order of formulas, tables, and numbers: correct those that can be recognized, keep the original image or re-enter what cannot be recognized. By handling it this way, the main text, formulas, and tables of the scanned paper can be preserved more completely.

Open OmniDoc to edit and export documents online