swair/prosophor/md2paper

Markdown to Academic Paper Converter

Convert Markdown files to properly formatted Word documents (and optionally PDF). Auto-detects language from filename (Chinese characters = CN format, otherwise EN/arXiv format).

When to Use

• User asks to convert .md files to .docx / Word / paper / PDF format
• User wants to generate academic papers, white papers, or theory documents from Markdown
• User references paper formatting standards (Chinese academic / arXiv style)

Prerequisites

pip install python-docx lxml docx2pdf

Usage

# Word only
python .claude/skills/md2paper/md2docx.py <input.md> <output.docx>
 
# Word + PDF
python .claude/skills/md2paper/md2docx.py <input.md> <output.docx> --pdf

arXiv Format Reference (learned from 2604.21804v1.pdf)

Title & Author Block

Phenomenological Detector Design and Optimization in Vertically-Integrated...
                                      (centered, normal weight, ~14-16pt)
 
Wonyong Chung                        (name)
Princeton University                 (affiliation)
Jadwin Hall, Washington Road, ...   (address)
wonyongc@princeton.edu               (email)
 
Qibin Liu                            (next author, same pattern)
SLAC National Accelerator Laboratory
...

Abstract Block

Abstract                                  ("Abstract" title case, bold, standalone)
We present the first implementation...    (flush-left, NO indent, single paragraph)

Body Text

1 Introduction                            (decimal section numbering)
High-energy physics (HEP) research...     (first-line indent, justified)
2.1 Bilevel optimization...               (hierarchical numbering)

Figure & Table Captions

Figure 1: Bilevel detector optimization... (full "Figure", bold colon separator, centered below)
 
Table 1: Optimization parameters...        (full "Table", bold colon separator, centered above)

Format Specifications

Chinese Format (paperol_cn_v2.0)

English Format (arXiv-style, from 2604.21804v1.pdf)

Page Layout

Abstract & Keywords Layout

Chinese

• 摘要： (bold label) + content paragraph, flush-left
• 关键词： (bold label) + keywords line, flush-left

English (arXiv)

• Header: Abstract (title case, bold, standalone line)
• Content: flush-left, NO left indentation, single paragraph
• Structure: inline bold keywords + text → "Context. ... Aims. ... Methods. ... Results. ... Conclusions. ..."
• Keywords: Key words. (bold) + keyword list, flush-left

Content parsing rules

• Mixed text + markdown tables → text paragraphs + real Word tables
• Empty lines separate text blocks
• Horizontal rules (---) skipped

Table/Figure Captions

Tables — caption ABOVE table:

• Chinese: 表1 名称
• English: Table 1: Name (full word "Table", bold colon separator)

Figures — caption below figure:

• Chinese: 图1 名称
• English: Figure 1: Name (full "Figure" word, NOT abbreviated "Fig.", bold colon separator)

Reference Citations

Inline [1], [2] etc. → blue superscript, Times New Roman.

Page Numbers

Centered in footer, PAGE field only.

Auto-processing

• YAML frontmatter: Parsed for author metadata, then stripped
• Author rendering: Title → Author (centered, NO prefix label) → content
• Author dedup: If plain-text author line matches YAML author, skipped
• Author fallback: If no YAML author, first non-empty line after title extracted as author
• Blockquote stripping: All > lines after title removed
• Footer removal: ## Software Citation / ## 软件引用 section and *Version:* / *Date:* / *Author:* footer stripped
• Language detection: Filename containing Chinese characters → CN format, else EN format

Supported Markdown

• Headings (# through ######)
• Fenced code blocks (```)
• Tables with header separator → real Word tables
• Unordered lists (-, *, +)
• Ordered lists (1., 2., etc.)
• Blockquotes (>)
• Inline bold (**) and italic (*)
• Images ([Image: alt])
• Horizontal rules (---)
• Reference citations ([1], [2])
• YAML frontmatter (metadata extracted, block stripped)