Undergraduate Thesis Word Formatter

Problem Addressed

Undergraduate thesis submissions often stall on Word formatting details rather than the manuscript itself: page size, margins, body font, line spacing, indentation, heading hierarchy, figure and table captions, page numbers, table of contents, and reference numbering can all require repeated manual fixes. This skill targets Chinese undergraduate thesis files in Word .docx and applies deterministic formatting based on common GB/T 7713 and GB/T 7714 conventions, making it useful as a pre-submission normalization step.

How It Works

• Page and style setup: format_paper.py configures A4 page size, default margins, Normal style, heading styles, and caption styles. The body text defaults to SimSun with Times New Roman, xiao-4 size, fixed line spacing, and first-line indentation.
• Heading detection: the script identifies headings by text patterns, such as “第一章”, “摘要”, “参考文献” as first-level headings, 1.1 as second-level headings, and 1.1.1 as third-level headings.
• Captions and TOC: it recognizes figure and table captions such as “图 1-1” and “表 1-1”, adds centered footer page numbers, and can insert a table-of-contents field for later generation in Word.
• References: format_references.py locates the references section, renumbers entries as [1] [2] [3], normalizes font, size, and line spacing, and checks for missing document type markers such as [M], [J], [D], as well as missing years.
• PDF export: docx_to_pdf.py supports docx2pdf with Microsoft Word or LibreOffice headless conversion, selecting an available method automatically.

Boundaries and Caveats

• Cover page: the skill does not design the cover page automatically; school templates, student IDs, names, and other cover fields still need manual handling.
• Split page numbering: Roman numerals for front matter and Arabic numerals for the main body usually require section breaks and per-section footer settings, which remain manual work.
• TOC update: the skill inserts a TOC field, but the final table of contents must still be generated by updating fields in Word.
• Reference parsing: the script is better suited to renumbering and checking existing references than fully parsing metadata such as authors, titles, and publishers from raw text.
• Heading accuracy: pattern-based detection can misclassify short list items or ordinary paragraphs as headings, so output should be reviewed manually.