论文排版与整理完全工作流 (Iterative OCR & Typesetting Pipeline)

Convert research-paper PDFs into citation-ready DOCX via OCR and OMML conversion.

42|6|Updated Apr 18, 2026
One-click install
npx skills add https://github.com/BoHeFan/academic-paper-writer-pro-2 --skill iterative-ocr-typesetting-pipeline
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: 论文排版与整理完全工作流 (Iterative OCR & Typesetting Pipeline)
Source: https://github.com/BoHeFan/academic-paper-writer-pro-2/tree/main/ocr_kb
Command: npx skills add https://github.com/BoHeFan/academic-paper-writer-pro-2 --skill iterative-ocr-typesetting-pipeline

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the problem of converting research PDFs (especially scanned or layout-heavy papers) into clean, structure-preserving DOCX that matches academic formats, without losing figures, tables, and equations.

Core Features & Use Cases

  • Iterative page OCR + structure normalization: extracts per-page text with strict heading-level hygiene and clean Markdown suitable for later typesetting.
  • Formula-to-OMML conversion workflow: forces all formulas into explicit LaTeX syntax so they can be converted into Word-native OMML reliably.
  • Figure/table detection with precise cropping: generates individually numbered figure images instead of embedding whole-page screenshots, preserving citation-ready visuals.
  • Incremental DOCX build with checkpointing and audit: appends each page immediately, supports suspension/continuation, and requires periodic verification every 2 pages.
  • Use Case: You need to re-typeset a conference paper in IEEE/APA format after correcting OCR issues, preserving global figure/equation numbering, and producing an editable DOCX for submission and revision.

Quick Start

Ask the agent to follow Pipeline A to iteratively OCR and typeset the provided PDF into a final DOCX in IEEE format, producing checkpointed intermediate outputs and figure-cropped assets.

Frequently Asked Questions about 论文排版与整理完全工作流 (Iterative OCR & Typesetting Pipeline)

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I convert a research paper PDF to IEEE format DOCX without losing equations and figures?

To convert a research paper PDF to IEEE format DOCX without losing equations and figures, apply page-by-page OCR with strict structure normalization, precise figure cropping, and LaTeX-to-OMML formula conversion. This process preserves global numbering and generates a citation-ready DOCX.

Can I extract figures from a PDF as individually cropped images instead of full page screenshots?

Yes, you can extract figures from a PDF as individually cropped images. The typesetting workflow detects figures and tables on each page, generating individually numbered image files rather than embedding full-page screenshots into the final DOCX document.

How do I convert LaTeX formulas to Word native OMML equations during PDF to DOCX conversion?

To convert LaTeX formulas to Word native OMML equations during PDF to DOCX conversion, the workflow forces all extracted formulas into explicit LaTeX syntax during OCR. This explicit syntax is then reliably converted into Word-native OMML equations in the final DOCX document.

Does incremental OCR with checkpointing support resuming long PDF to DOCX typesetting workflows?

Yes, incremental OCR with checkpointing supports resuming long PDF to DOCX typesetting workflows. The pipeline appends each page immediately, saves deterministic intermediate artifacts, and supports suspension and continuation without losing previous processing progress.

What is the best way to handle OCR errors when re-typesetting academic papers into APA or Springer templates?

The best way to handle OCR errors when re-typesetting academic papers into APA or Springer templates is to use a pipeline with audit checks and a needs_review resolution step. It ensures the final DOCX is only produced after verifying intermediate outputs every two pages.

Why does my PDF to DOCX academic typesetting workflow lose global counters for figures and tables?

PDF to DOCX academic typesetting workflows lose global counters for figures and tables when lacking strict structure normalization. An iterative page-by-page pipeline maintains global counters for figures, tables, and equations throughout the entire conversion process.