Chinese Word Typo Checker

Problem Being Solved

Chinese Word document proofreading often stalls on locating the exact error, choosing the correct fix, and proving that formatting survived. Manual review can miss homophones, similar-looking characters, dropped or duplicated words, while blind replacement may damage proper nouns, technical terms, or context-specific phrasing. This skill targets .docx / .doc files with a controlled workflow for Chinese typo correction.

How It Works

It extracts readable text from body paragraphs, tables, headers, footers, and text boxes using the bundled scripts. The model then identifies homophones, similar-shaped typos, common character substitutions, missing or duplicated words, and idiom typos. Findings are stored as a structured list with fields such as wrong, correct, reason, context, and occurrence, keeping the revision pipeline inspectable.

The skill writes corrections through typo_fixer.py using Word-native tracked changes: deleted text, inserted text, and a comment explaining the reason for each change. When the same wrong token appears multiple times but only one location should be fixed, context can anchor the paragraph text, while occurrence can select the Nth occurrence; the two can be combined. The output is named OriginalName_CheckOut_timestamp.docx and placed next to the input file.

After revision, verify.py enforces checks: paragraph counts match, accepting all revisions produces only the expected changes, and the package structure remains intact. If validation fails, the document must not be released; the typo list should be corrected and the fix-and-verify loop retried, up to 3 times.

Boundaries

It only handles Chinese character typos, not grammar, word order, collocation, layout, styling, or English spelling. Input must be a Word document. .doc conversion relies on Windows, Microsoft Word, and pywin32. Footnotes and endnotes are outside the current extraction scope, so documents containing them should be flagged for possible unchecked areas. For uncertain phrasing, prefer a comment suggestion over a forced replacement.