OmniDocHow to extract thesis document content: Zotero workflow
Researchers often save PDFs of papers in Zotero, but what they really need is the text, formulas, and tables. This article explains how to distinguish the text version from the scanned version, and use OmniDoc desktop to complete extraction, translation and verification in the Zotero document workspace.
Key points of this article
- First determine whether the PDF is a text version or a scanned version. The two are processed in different ways. Choosing the wrong one will result in missing or garbled text.
- Organize the entries and PDF attachments in Zotero first, and then extract, translate and export them to the correct paper.
- The scanned PDF does not have optional text and needs to be OCR recognized before translation or export.
- Extraction results need to be spot-checked for formulas, tables, references and terminology, and machine translation and OCR results need to be checked against the original text.
Many scientific researchers' documents are stored in Zotero, and the PDF of each paper is hung under the corresponding entry. But what is really needed is often the text, charts, formulas and references in the PDF. How to stably extract these contents is a very specific problem. Copying and pasting directly often loses paragraphs and formulas, and manually transcribing page by page is too slow.
This article is written for graduate students, researchers, and literature review writers who need to deal with PDFs of their dissertations. This article will first teach you how to determine the type of PDF, then explain how to use the OmniDoc desktop to find attachments, translate or identify content in documents organized by Zotero, and finally give methods to check the results. After reading, you can follow the steps to extract the content of a paper into a readable and editable file.
How to choose software to extract document content? One sentence conclusion
First determine the type of PDF, and then select the extraction method: the text version of PDF can read text directly, while the scanned version of PDF needs to be OCR recognized first. Both types of files should retain the original PDF in Zotero and do not overwrite the original text with the processed file.
| your situation | Recommended practices | reason |
|---|---|---|
| Text version PDF, only the text text needs to be translated | Select translation output only in OmniDoc and check after downloading | Text can be read directly, and the output results are cleaner. |
| Text version PDF needs to be compared paragraph by paragraph | Select bilingual comparison output | Original text and translation are side by side, making it easy to check terms and long sentences |
| Scanned PDF, cannot be selected when selecting text | Use the OCR function of the PDF editor to recognize and then process, or directly upload the translation | There is no optional text in the image, it needs to be recognized first |
| Need to continue editing the table or text | Export to Excel or Word after recognition | After exporting, it can be modified in commonly used office software. |
| Just need a reference list | Mainly based on bibliographic data in Zotero, compared with PDF references item by item | Bibliographic data is already in Zotero, making verification more efficient |
Before you start: Confirm these points first
Taking a few minutes to confirm the following items before starting can avoid repeated rework later.
- Confirm PDF type. Try dragging and selecting a section of text in the reader. If you can select text, it means it is a text version; if you cannot select, you can only select a whole page, usually a scan or picture.
- Confirm that the attachment is the original PDF. The files automatically crawled by Zotero are sometimes just summary pages or incomplete versions. Check the page number and author information, and confirm that the attachments correspond to the entries.
- Confirm whether you want to keep formulas and tables. The formulas, tables and legends in the paper have complex structures. Before extracting, look at several samples to determine whether the output meets your needs.
- Confirm the scope of use of the document. Only deal with documents that you have the right to use for personal study, organization and citation.
- Allow time for verification. Both machine translation and OCR may have errors. Important paragraphs, values and terms must be confirmed against the original text.
Steps to extract paper content through Zotero on OmniDoc desktop
The following process is suitable for batch sorting of papers under a research topic. When working with only a single file, you can also upload PDF directly to omnidoc.top.
- Organize literature in Zotero. Create an entry for each paper, confirm that the title, author, and year are complete, and make sure that the PDF attachment is hung below the entry.
- Download and install the desktop client. Visit download.omnidoc.top and select the Windows, macOS or Linux version to install according to the system.
- Open the Zotero literature workspace. Enter the literature workspace connected to Zotero in OmniDoc desktop, load your literature library, and list the items that need to be processed.
- Complete missing PDFs. For entries without attachments, use the Find PDF function in the workspace to obtain the file, and check whether the title and author are consistent with the entry.
- Select the output method and translate. Select a paper, select bilingual comparison or translation only according to your needs, and start translation. Formulas, tables, and images are retained along with the layout.
- Process the scanned version of the document. If the PDF is a scan, use the OCR function in the PDF editor to recognize the text before translating or exporting.
- Export editable files. If you need to continue to modify the table or text, export to Word or Excel; after exporting, check whether the paging and table are complete.
- Save the results and make a record. Save the translation or export file to the project folder, and write down the processing date and output method in the notes of the Zotero entry for easy tracing later.
Check each item after completion
After getting the results, do not quote or transfer them directly, but first check according to the table.
| Check items | How to see | What to do if there is a problem |
|---|---|---|
| Is the content complete? | Compare the number of pages, turn to the last page, and make sure that the references and appendices are in | If a page or segment is missing, go back to the original PDF and reprocess it, or try again after switching to OCR. |
| Formula | Randomly select three to five formulas and compare them with the original text one by one | If the formula is confused, the original text shall prevail and will be used after manual correction. |
| Forms and annotations | Check whether the number, values and units of rows and columns are consistent | If the values are inconsistent, go back to the original picture and check it. Don't modify it according to your impression. |
| Professional terminology | Whether the translation of the same term is unified in the full text | Establish a glossary and replace it uniformly to avoid multiple statements of the same concept. |
| Zotero association | Whether the entry still points to the correct PDF attachment | Delete the error attachment, re-attach the correct file, and check the author and year |
# frequently asked questions (FAQ)
1. Can all the software that extract the contents of the literature handle the scanned version?
No. The text in the scanned version of PDF is an image and cannot be selected directly. It needs to be recognized by OCR first. The PDF editor on the desktop side of OmniDoc provides OCR function, and the web side can also upload a scanned version of PDF directly for translation.
2. Is there a free extraction tool available?
When choosing a free tool, it is recommended to look at several aspects: file size and page limit, whether the exported file is watermarked, whether the privacy policy describes how the file is stored and deleted, and whether the exported typesetting, formulas and forms are retained. When the paper contains a large number of formulas and forms, first use a short article to do the test, and then decide whether to batch processing.
3. Can I use it without Zotero?
Sure. Upload a single PDF in omnidoc.top, select bilingual control or translation only, and download the results. However, the collation of literature entries and the search and connection of PDF need to be completed in the Zotero document workspace on the desktop side.
4. Can the extracted translation be written directly into the citation of the paper?
Direct citation is not recommended. The translation needs to be checked against the original, especially numerical values, negative words and technical terms. When quoting, the original text shall prevail, and the source of the original text shall be indicated in accordance with the specification.
5. Can formulas and references be retained?
The translation output retains the layout of formulas, tables, and pictures. The reference list is recommended to be compared with the bibliographic data in Zotero one by one to confirm that the author, year, journal name and volume page number are the same.
# Summary
The key to the content extraction of the paper is to first judge the type of PDF, and then keep the entry corresponding to the attachment in the Zotero. After processing according to the steps in this article, check the formulas, tables and terms item by item, and the available literature content can be obtained. Zotero workflow makes the collation and extraction of multiple papers more organized and easy to trace in the future.
