OmniDocHow to retain the original format after table recognition
After table recognition in scanned documents or pictures, merged cells, borders, and number formats are often lost. This article explains what format preservation means, how to choose the output format according to the purpose, and how to check item by item after exporting.
Key points of this article
- Format preservation means that the row and column structure, merged cells, number format and alignment are still there, not just the text content is recognized.
- If you want to continue calculating the data, export it to Excel; if you want to format it for printing or give it to others to read, it is more reliable to export it to Word.
- The scanned document must be OCRed first and then exported as a table. Otherwise, the output is usually just a full page image or scattered text.
- After exporting, check whether the merged cells, decimal places, cross-page headers and numbers are misaligned, and correct them promptly if any problems are found.
- The identification results may be wrong, and the numbers involving the amount, date, and serial number must be checked one by one with the original.
After many people use the recognition function to transfer tables in contracts, financial statements or papers, they find that the text is correct, but the table is messed up: the merged cells are split, the numbers are strung into other columns, and the borders and alignment disappear. Manual re-adjustment often takes more time than the identification itself.
This article is for readers who need to convert scans, images, or tables in PDF into editable files. After reading, you will know what format preservation means, how to choose the output format according to subsequent uses, and what to check after exporting.
What I’m talking about here is the OCR and PDF export methods in the OmniDoc desktop client, and it’s also a general idea that applies to other tools.
Should the format be retained after table recognition? One sentence conclusion
Whether or not the format should be kept intact depends on what you want to do after you get the file. The table below provides recommended practices for common situations.
| your situation | Recommended practices | reason |
|---|---|---|
| To sum, filter, and pivot numbers | Export to Excel and check each column of numbers | Only after the row and column structure is retained can it be calculated directly. |
| To print or send to others to read | Export to Word, check layout and table borders | The layout of text and tables is closer to the original |
| Only the text content in the table is required | Copy text or export plain text and organize it yourself | Don’t pursue layout when content is your priority |
| There are a large number of merged cells in the table | After exporting, check the merged areas one by one and correct them manually if necessary. | Merging cells is an error-prone part of identification |
| The original text is bilingual or needs translation | Using Document Translator, choose output that preserves tabular structure | Tables, formulas and pictures can still be retained after translation |
Before you start: Confirm these points first
First confirm whether the document is a scanned copy. If the PDF is directly exported from an electronic document, table text can usually be selected; if only the entire page image can be selected, it means that it is a scan or picture and needs to be OCRed first. The judgment method is very simple: try to use the mouse to drag and select a piece of text in the table in the reader.
Find out what "format preservation" means. It contains at least four categories of content: row and column structure (how many rows and columns does the table have, which cells are merged), location of cell content (in which column the number is, which row the title is in), number format (decimal places, thousandths, percent signs), and borders and alignment. OCR itself is only responsible for recognizing text and approximate location. Structural restoration depends on the export process, so both processes must be confirmed.
Check the original for clarity in advance. Blurred, skewed, and shadowed scans will cause inaccurate recognition of cell boundaries, resulting in column misalignment. If the original is too poor, try rescanning it first or using a clearer version instead of fixing it later.
Determine which numbers must be verbatim. If you identify the wrong information such as amount, date, serial number, and tax ID number, it will cause subsequent trouble. No matter how neat the recognition results are, they need to be compared with the originals.
- prepare the target software. * if your follow-up work is done in desktop office software, choose the corresponding export format; if you want to continue editing online, you can put the results in the UniDoc Page Editor (app.unidoc.top).
# Export steps with table structure in OmniDoc desktop client
- download and open the desktop client. download the appropriate version of your system from download.omnidoc.top, open it after installation, and load the PDF files that need to be processed.
- determine whether OCR is needed. if the text cannot be selected in the table on the page, it means that this is a scanned piece. You need to execute OCR first to make the text recognizable.
- execute OCR and wait for completion. text recognition of the page on which the form is located or the entire document. When there are a large number of pages, you can process only the pages that contain the table, reducing the waiting time.
- read the recognition results page by page. Open the form page before export and check whether the rows and rows are aligned and whether the title and header are in the correct position. When you find an obvious error, you can modify the paragraph text directly in the document.
- Select the export format. choose to export to Excel when you need to continue calculation, and export to Word when you need typesetting or printing. Both formats can be done in the same PDF editing process.
- specify the save location and export. give a distinguishing name to the result document, such as "contract schedule _ identified version", to avoid overwriting the original.
- Open the exported file. View the table in the target software, not just the preview. Focus on the items listed in the next section.
- make a second arrangement if necessary. * if there is still a deviation in the table structure, you can manually adjust the merged cell or supplementary header in the corresponding editor, and then save it as the final version.
# check item by item after completion
Do not rush to use it after export, check it item by item according to the table below.
| Check item | What do you think? | What if there's a problem? |
|---|---|---|
| Number of rows and columns | Number of rows and columns compared with the original | Go back to the identification page, check for blank lines or split cells, and merge or split manually if necessary |
| Merge cells | Look for the headings that span rows and columns in the original, and see if it is still a whole block after export. | Re-merge in the target software and confirm that the merged content has not been deleted |
| Numbers and decimal places | Spot check amount, percentage and position of decimal point | Compare with the original one by one, pay special attention to the misrecognition of 0 and O, 1 and l |
| Whether the numbers are concatenated | See if the numbers fall into the correct column. | Usually the form line is not recognized correctly. Move it manually after returning to the original to confirm. |
| Spread header | When the table spreads, check whether there is a header on the second page. | Fill in the header on each page, or unify the header row in the target software |
| Text content | Date, number and name of spot check | Check all the key fields, not just one or two spot checks |
| Border and alignment | Check whether the frame is complete and whether the text is centered or left against the original. | The missing frame can be added in the target software, and the alignment can be set uniformly. |
The recognition result is the machine output, the important content must be checked with the original item by item before using.
# Core comparison
Several common output methods have their own emphasis, which can be selected according to the subsequent use.
| Output mode | Row and column structure | What is suitable for doing? | Main risk |
|---|---|---|---|
| Excel | Keep it intact and easy to calculate | Data analysis, summary, screening | Typesetting will deviate from the original when there are more text paragraphs. |
| Word | Keep the table frame, the layout is close to the original | Print, file and send it to others to read | Complex merged cells may need to be corrected manually |
| Pure text | Structure loss | Just the content itself. | Numbers are easy to serial, but it is difficult to restore tables. |
# which method is suitable for you?
- readers doing data statistics : give priority to exporting Excel, focus on checking number columns and decimal places, and do not rush typesetting. - readers who need to archive or print : priority is given to exporting Word, focusing on table borders, headers, and spread effects. - readers who only care about the content : keep the text, but the key numbers should still be compared with the original.
# frequently asked questions (FAQ)
1. Is there a shortcut key for form recognition?
This article describes the export format and check steps, and does not rely on special shortcuts. Copy and paste can be done with the keyboard shortcuts commonly used by the system, but whether the format can be retained depends mainly on the choice of the output format rather than pressing the key.
2. Why is there a missing line in the form in the scan?
The grayscale, shadow and tilt during the scan will make the table lines lighter or even broken, and the adjacent cells may be mistaken for the same grid when identifying. Rescan more clearly, or fill up the border after export, the effect is usually significantly improved.
3. What if the number becomes text after exporting Excel?
This situation is often seen in the recognition of numbers as words. You can format the column as a number in the target software, and then check that the decimal point is correct. After the number is transformed into text, it can not directly participate in the summation, but needs to be converted first.
4. The form spans several pages, how to keep the header consistent?
Check that each page has a header before exporting. If there is only the first page, you can add the header row in the target software. In this way, when subsequent filtering and printing, each page can understand the meaning of the column.
5. Can the format be preserved completely without the right software?
If the original structure is particularly complex, any automatic recognition may require manual correction. A more practical approach is to accept the automatic results and then manually adjust the key parts with the target software, rather than expecting a single export to be completely consistent.
# Summary
The key to whether the original format can be retained after the form recognition is to judge whether it is a scanned piece, then select Excel or Word export according to the purpose, and finally check the rows and rows, merge cells and numbers one by one. Following these three steps, most tables can get results that can be used directly.
