OmniDocWhat are the formula identification methods? Comparison of several common ways
There are four ways to recognize formulas: desktop software, online web tools, OCR in document editors, and manual input. This article explains based on the file source and output format what situations each type is suitable for, as well as the cost of choosing the wrong one.
Key points of this article
- Electronic PDFs whose formulas can be selected are directly extracted first, without the need for image recognition.
- Scans and photos need to be OCRed first. After recognition, the superscripts, subscripts and symbols must be checked against the original images.
- Whether the output is to be LaTeX or editable Word equation objects determines the path chosen.
- When there are only a few simple formulas, it is often more time-consuming to enter them manually than to modify them after identifying them.
- Online tools look at file size limits, watermarks and privacy instructions before deciding whether to upload.
If you have a paper, a handout, or a scanned test paper, and the formulas in it cannot be copied or modified, you will encounter the problem of formula recognition. Many people search for "what are the formula identification methods", but what they actually want to know is: it also turns the formula into editable content, what situations each of the several routes is suitable for, and whether there will be rework if the wrong choice is made.
This article does not list tools, nor does it teach the specific operations of each software. Instead, it classifies common methods into four categories: desktop software, online web tools, OCR in document editors, and manual re-typing. The conditions, prices, and applicable scenarios of each category are different. Understanding these differences is more useful than remembering a product name.
After reading it, you should be able to judge two things: which type of file you have, and what format the required output is. After making a clear judgment, it will not be easy to take detours when choosing a path.
What are the formula identification methods? One sentence conclusion
One sentence conclusion: first look at the file source, and then look at the output format. When the file itself has a text layer, it is preferable to use a document tool to directly extract it; only when it is a picture or a scan, OCR recognition is required; when there are few formulas or the structure is complex, manual entry is more stable.
| your situation | Recommended practices | reason |
|---|---|---|
| Electronic PDF, formulas can be selected | Copy directly, or use a PDF editor to export to Word and check | The formula already has a text layer and does not require image recognition. |
| Scanned copy or photo taken with mobile phone | OCR first, then check the formulas one by one | There are no characters in the picture that can be directly extracted |
| Requires LaTeX source code or editable formula object | Input or convert using a document editor that supports formula editing | The output format determines whether the results can be used directly |
| There are only two or three simple formulas | Manual entry | It still needs to be modified after recognition, it is faster to write directly |
| The entire information should be converted into an editable document | Use the desktop client to process it as a whole, and then check it page by page. | It is easier to deal with the entire article than to deal with the formulas one by one. |
What is formula identification?
Formula recognition is to convert the image or layout information of the formula into a representation that can be edited and saved by the computer. Commonly used ones include LaTeX code, MathML, and formula objects in Word. The difference between it and ordinary text recognition is that formulas have two-dimensional structures such as superscripts and subscripts, fractions, radicals, and matrices. If read as linear text, the structure can easily be disrupted.
For example, a fraction like $\frac{a}{b}$ may be recognized as "a/b" by ordinary OCR, which looks like text, but the structure of the fraction line has been lost. What formula recognition needs to do is to restore symbols and their positional relationships together.
How it works
The processing paths for files from different sources vary greatly.
- Formula in electronic PDF: The file itself saves character and position information. When extracting, it mainly restores the structure, which is less difficult. Being able to use the mouse to select characters in the formula usually indicates this category.
- Formulas in scans or pictures: You need to find the symbols and layout through OCR first, and then reconstruct the formula structure. Scan quality, resolution and tilt angle all affect the results.
- Handwriting formula: Handwriting recognition is more affected by handwriting, and the results of handwritten manuscripts usually require more careful manual verification.
Common misunderstandings
Misunderstanding 1: Once OCR recognizes it, it can be used directly. Common errors in formula recognition include incorrect placement of superscripts and subscripts, misrecognition of Greek letters, incorrect bracket levels, and missing decimal points. The recognition results must be compared with the original images one by one, especially papers and examination materials. If one symbol is wrong, the entire derivation will not be valid.
Misunderstanding 2: Exporting to Word must be a formula object. Some conversions will only turn the formula into an image or plain text, which can no longer be edited. The judgment method is very simple: double-click the formula after exporting to see if it can enter the formula editing state. Only if it can be entered can it be considered truly editable.
Misunderstanding 3: The principles of all approaches are the same. Functionality, processing scope and output formats vary within online tools, desktop software and document editors. When choosing a path, you should look at what inputs it supports and what the output is, rather than just "whether it can be recognized".
📊 Core comparison
| way | suitable document | Output characteristics | Main cost |
|---|---|---|---|
| Desktop software (including PDF editor) | Electronic version of PDF, scanning parts | In local processing, it can be exported to Word, Excel and other formats | Need to install, be familiar with the file processing process |
| Online web tools | Pictures, small files | Open and use, and the result is mostly LaTeX or text | Pay attention to privacy policies and file size restrictions before uploading |
| OCR within the document editor | Scan the document and need to edit the whole article | Modify it directly in the document after identification. | The complex formula still needs to be modified manually. |
| Re-enter manually | Few formulas and simple structure | Completely controllable | Time-consuming and easy to copy wrong |
When comparing, you can ask yourself three questions: whether the file is electronic, whether the result is picture, text or editable formula, and whether you are dealing with several formulas or the whole document. The answers to the three questions basically determine the path.
# how to use it
If you have a scanned copy and want to export to an editable Word document, you can do the following.
- Confirm the file type first. In PDF, try to select a text near the formula, but if not, it usually means that it is a scanned piece or picture.
two. Scanned pieces can be uploaded, translated or processed directly. OCR the file in the PDF editor of the OmniDoc desktop client, and check the formula area when you are finished. The client download address is download.omnidoc.top.
- After completing the OCR, export the document as Word so that both the formula and the text can be modified in Word.
- Check the formula one by one in Word: whether you can enter the editing state, whether the superscripts, fractions and Greek letters are consistent with the original drawing.
- If you need LaTeX source code, or if you want to collaborate online, you can put the content into the UniDoc Page Editor (app.unidoc.top) to continue processing, which supports formula editing and can be transferred between Word and LaTeX.
The results of the OCR need to be checked manually, especially the derivation in the paper, the values in the exam and the terms of the contract. When checking, put the result side by side with the original picture, and look at the formula one by one, it is easier to find errors than to read through.
# which method is suitable for you?
-the paper is an electronic version of PDF, can select the text: directly extract, and then check the formula one by one. Handout taken by mobile phone or scanned test paper: do OCR first, and then modify the formula structure manually. -only two or three formulas need to be written into the document: enter them manually, usually faster than after identification. -the whole document should be converted into an editable document: processed as a whole with desktop software, and then checked uniformly.
# frequently asked questions (FAQ)
1. Is formula recognition the same thing as OCR?
Not exactly. OCR is a general technology for text recognition, and formula recognition is one of the more difficult ones, which needs to restore the two-dimensional structure between symbols. Ordinary OCR will treat formulas as ordinary words, so superscripts and fractions are easy to be confused.
2. How to choose the free formula recognition method?
You can look at four points when choosing: file size limit, whether the output is watermarked, the privacy description after upload, and whether typesetting and formulas can be retained. If you only look at the word "free", it is easy to rework on the results.
3. Is the online formula recognition method safe?
It depends on the sensitivity of the document. You can use online tools to disclose information; for unpublished papers, contracts or internal materials, it is recommended to read the privacy policy first, or give priority to desktop software processed locally.
4. How to judge whether the recognition result is correct or not?
Compare the results side by side with the original picture, focusing on fractions, superscripts, Greek letters, parenthesis pairs and decimal points. If you get LaTeX, you can render it to see if the display is consistent with the original.
# Summary
There is no universal way of formula recognition, the key is to determine whether the document is an electronic version or a picture, and then decide which method to use. No matter which route the result comes from, it should be checked with the original text one by one, especially those involving derivation and numerical values.
