name	paper-translator-korean
description	Translate academic papers and technical documents in markdown format from any foreign language to Korean. Use when user requests translation of research papers, academic documents, technical books, or scientific articles to Korean.

Paper Translator (Any Language to Korean)

When to Use

Use this skill when:

User requests translation of academic papers or technical documents to Korean
Input is markdown-formatted academic content in any language (English, German, French, Spanish, Japanese, Chinese, etc.)
User asks to translate research papers, journal articles, technical books, or scientific documents
Source language can be any foreign language - the skill will automatically detect and translate to Korean

Default behavior: Full paper translation in one go

Translate the entire paper automatically without user approval
Process all sections sequentially without asking for confirmation
DO NOT summarize - translate every single word and sentence completely

Translation Guidelines

1. Pre-processing: Clean OCR Artifacts

IMPORTANT: Papers converted from OCR may contain noise that should be ignored:

Meaningless line breaks: Remove unnecessary line breaks within paragraphs
- Example: "This is a\nsentence." → "This is a sentence."
- Keep intentional paragraph breaks
Header/Footer artifacts: Ignore and remove:
- Page numbers (e.g., "Page 5", "5", "- 5 -")
- Journal names repeated on every page
- Author names in headers
- Copyright notices (e.g., "© 2023 IEEE")
- Conference names (e.g., "NeurIPS 2023")
- DOI strings at top/bottom
- Running headers/footers
Hyphenation errors: Merge words split across lines
- Example: "compu-\nter" → "computer"
Extra whitespace: Normalize spacing between words and paragraphs
OCR misreads: Use context to correct obvious errors if possible

Clean first, then translate: Remove all OCR noise before beginning translation.

2. Analyze the Paper Structure

Auto-detect source language: The document can be in any language (English, German, French, Spanish, Japanese, Chinese, etc.)
Identify sections (Abstract/Zusammenfassung/要約, Introduction/Einleitung/序論, Methods, Results, Discussion, Conclusion, etc.)
IMPORTANT: Identify sections that should NOT be translated:
- References/Bibliography (REFERENCES/Literatur/参考文献, Bibliography, Works Cited, etc.)
- Appendix/Appendices (APPENDIX/Anhang/付録, Supplementary Material, etc.)
- Acknowledgements/Acknowledgments (ACKNOWLEDGEMENTS/Danksagung/謝辞, Acknowledgment, etc.)
- These sections should be copied as-is without translation
Note technical terms, equations, citations, and figures (in any language)
Preserve markdown formatting (headers, lists, tables, code blocks)

3. Translation Strategy

CRITICAL - Complete Translation Without User Approval:

Translate the entire paper automatically from start to finish
NO summarization - translate every single word and sentence completely
NO user confirmation - process all sections sequentially without asking
Start with Abstract and continue through all sections automatically
Only stop at sections that should NOT be translated (References, Appendix, Acknowledgements)

Automatic Full-Paper Approach:

Analyze the paper structure silently (no user prompts)
Start translating from Abstract automatically
Continue through all sections without stopping for approval
Save all translated sections progressively to the output file
Complete the entire translation in one continuous workflow

Translation Guidelines per Section:

Complete Translation: Translate every single sentence - never summarize or skip content
Academic Terms: Maintain original language terms in parentheses for key technical vocabulary
- Example (English): "machine learning (머신러닝)" on first occurrence
- Example (German): "Softwarearchitektur (소프트웨어 아키텍처)" on first occurrence
- Example (Japanese): "機械学習 (머신러닝/기계학습)" on first occurrence
Maintain Formality: Use formal Korean academic writing style (합니다체)
Preserve Citations: Keep citation formats unchanged [1], (Smith et al., 2023), etc.
Equations: Keep mathematical notation as-is
Figures/Tables: Translate captions and descriptions to Korean

4. Step-by-Step Translation Process

CRITICAL: Automatic full-paper translation without user approval!

Clean OCR artifacts (FIRST STEP):
- Remove page numbers, headers, footers
- Fix meaningless line breaks within sentences/paragraphs
- Merge hyphenated words split across lines
- Remove repeated journal/conference information
Analyze structure silently:
- Identify all sections in the paper
- Note sections to skip (References, Appendix, Acknowledgements)
- DO NOT ask user for confirmation - proceed automatically
Translate ALL sections automatically:
- Start from Abstract (or equivalent in source language)
- Continue through all sections sequentially without asking
- SKIP sections that should NOT be translated:
  - References/Bibliography: Translate section header to "## 참고문헌 (REFERENCES/Literatur/参考文献)" but keep all reference entries in original language
  - Appendix: Translate section header to "## 부록 (APPENDIX/Anhang/付録)" but keep content in original language
  - Acknowledgements: Translate section header to "## 감사의 글 (ACKNOWLEDGEMENTS/Danksagung/謝辞)" but keep content in original language
- NEVER ask user for approval between sections
For each section translation:
- Complete translation: Translate every single word and sentence - no summarization
- Maintain section header format
- For technical terms:
  - First occurrence: "English term (한국어 번역)"
  - Subsequent: Use Korean or English based on common usage
- Preserve all markdown formatting:
  - Headers (#, ##, ###)
  - Lists (-, *, 1.)
  - Code blocks (```)
  - Tables (|)
  - Links and references
Save progressively:
- Save translated content as each section completes
- Continue to next section automatically
- Report completion only when entire paper is done

Example workflow:

User: "이 논문을 번역해줘" (또는 "ml_survey.md를 번역해줘")
Claude: [Silently analyzes structure] → [Translates Abstract completely] → [Saves to ml_survey_ko.md with YAML header] → [Translates Introduction completely] → [Appends to ml_survey_ko.md] → [Continues through all sections] → "전체 논문 번역이 완료되었습니다. ml_survey_ko.md 파일에 저장했습니다."

4. Quality Checks

Ensure natural Korean sentence flow
Verify technical accuracy
Check that all sections are translated
Confirm markdown structure is preserved

5. File Saving Protocol

CRITICAL: Always save translated sections to a file with YAML header

File naming convention:

Original file: paper.md or research_paper.md
Translated file: paper_ko.md or research_paper_ko.md
Pattern: {original_filename}_ko.md

YAML Header Requirement:

ALWAYS prepend YAML header from header.yaml before any content
This is critical for proper HTML rendering with embedded images
The header contains embed-resources: true for image embedding
Header path: /media/restful3/data/workspace/paperflow/header.yaml or header.yaml in project root

Saving behavior:

First section (Abstract): Create new file with YAML header
- Read header.yaml from project root
- Write YAML header to new {filename}_ko.md file
- Append translated Abstract section
- Do NOT write YAML header again in subsequent sections
Subsequent sections: Append only the translated content
- Append to existing {filename}_ko.md (which already has YAML header)
- Do NOT add YAML header again
File location: Save in the same directory as the original file

Example workflow:

Original: /papers/ml_survey.md

Step 1: Translate Abstract
→ Create /papers/ml_survey_ko.md with:
   1. YAML header from header.yaml
   2. Translated Abstract content

Step 2: Translate Introduction
→ Append Introduction to /papers/ml_survey_ko.md (no YAML header)

Step 3: Translate Methods
→ Append Methods to /papers/ml_survey_ko.md (no YAML header)

Final result: Complete translated paper in ml_survey_ko.md with YAML header

After each section translation:
- Save/append the section to {filename}_ko.md silently
- Continue to next section automatically without asking
- Only report when entire paper is complete
If no original filename provided:
- Ask user: "번역 결과를 저장할 파일명을 알려주세요 (예: paper_ko.md)"
- Or use default: translated_paper_ko.md
If header.yaml is not found:
- Use a minimal fallback header with embed-resources: true
- Warn the user: "header.yaml을 찾을 수 없어 기본 헤더를 사용했습니다."

Examples

Example 1: OCR Noise Cleaning

Input (with OCR artifacts):

NeurIPS 2023                                                    Page 5

## Abstract
Machine learning has revolu-
tionized data analysis...

© 2023 Neural Information Processing Systems

After Cleaning:

## Abstract
Machine learning has revolutionized data analysis...

Example 2: Translation with File Saving (with YAML Header)

Scenario: User has deep_learning_survey.md

Claude's workflow:

1. Analyze structure
   → Sections: Abstract, Introduction, Related Work, Methods

2. Translate Abstract (FIRST SECTION - ADD YAML HEADER)
   → Read header.yaml from project root
   → Create deep_learning_survey_ko.md with:
   ---
   lang: ko
   format:
     html:
       toc: true
       embed-resources: true
       theme: cosmo
   ---

   ## 초록 (Abstract)
   딥러닝(deep learning)은...

3. Translate Introduction (SUBSEQUENT SECTION - NO YAML HEADER)
   → Append to deep_learning_survey_ko.md:
   ## 서론 (Introduction)
   최근 몇 년간...

4. Continue for each section
   → Final file: deep_learning_survey_ko.md (complete translation with YAML header)

Example 3: Translation with Technical Terms

Input (English):

## Abstract
Machine learning has revolutionized data analysis...

## Introduction
Recent advances in deep learning...

Output (Korean) saved to {filename}_ko.md:

## 초록 (Abstract)
머신러닝(machine learning)은 데이터 분석에 혁명을 가져왔습니다...

## 서론 (Introduction)
최근 딥러닝(deep learning)의 발전은...

Input (German):

## Kapitel 4. Architektonische Merkmale definiert
Softwarearchitekten müssen architektonische Merkmale definieren...

## Architektonische Merkmale und Systemdesign
Um als Architekturmerkmal zu gelten, muss eine Anforderung drei Kriterien erfüllen...

Output (Korean) saved to {filename}_ko.md:

## 4장. 아키텍처 특성 정의 (Kapitel 4. Architektonische Merkmale definiert)
소프트웨어 아키텍트는 아키텍처 특성(architektonische Merkmale)을 정의해야 합니다...

## 아키텍처 특성과 시스템 설계 (Architektonische Merkmale und Systemdesign)
아키텍처 특성으로 간주되려면 요구사항이 세 가지 기준을 충족해야 합니다...

Example 4: Handling References Section

Input (English):

## VI. CONCLUSION
This paper reviewed major AI frameworks...

## REFERENCES

[1] J. Smith, "Machine Learning Basics," IEEE Trans., 2023.
[2] A. Jones, "Deep Learning Systems," ACM Conf., 2024.

Output (Korean) - References kept in English:

## VI. 결론 (CONCLUSION)
본 논문은 주요 AI 프레임워크를 검토했습니다...

## 참고문헌 (REFERENCES)

[1] J. Smith, "Machine Learning Basics," IEEE Trans., 2023.
[2] A. Jones, "Deep Learning Systems," ACM Conf., 2024.

Special Considerations

OCR Artifact Patterns to Remove

Common patterns found in OCR-processed papers:

Page numbers: Page 1, - 1 -, 1 of 10, standalone numbers at top/bottom
Running headers: Author names, paper titles repeated on each page
Journal info: IEEE Transactions on..., Proceedings of...
DOI: DOI: 10.xxxx/xxxxx at top or bottom
Timestamps: Received: Jan 2023, Accepted: Mar 2023
Hyphenation: compu-\nter, analy-\nsis, informa-\ntion

Content Preservation

For mathematical papers: Keep all LaTeX/equations unchanged
For CS papers: Programming terms often stay in original language (English, etc.)
For medical papers: Use established Korean medical terminology
For multilingual papers: Preserve technical terms in their original language when appropriate
Maintain all reference numbers and bibliography format (regardless of source language)

Common Technical Terms Guide

Computer Science & AI

Machine Learning → 머신러닝
Deep Learning → 딥러닝
Neural Network → 신경망
Algorithm → 알고리즘
Dataset → 데이터셋
Model → 모델
Training → 학습
Validation → 검증
Testing → 테스트
Accuracy → 정확도
Precision → 정밀도
Recall → 재현율

Research Methodology

Methodology → 방법론
Experiment → 실험
Analysis → 분석
Results → 결과
Discussion → 논의
Conclusion → 결론
Hypothesis → 가설
Variable → 변수
Correlation → 상관관계
Statistical Significance → 통계적 유의성

Translation Best Practices

Complete translation without summarization (MOST IMPORTANT): Translate every single word and sentence
Automatic workflow: Process entire paper without user approval between sections
First sentence of each section: Translate completely, setting the tone
Technical accuracy over literal translation: Prioritize meaning
Consistency: Use the same translation for repeated terms
Natural flow: Korean sentence structure may differ from English
Academic conventions: Follow Korean academic writing norms

Handling Token Limits

Error: "Claude's response exceeded the 32000 output token maximum"

This means the paper is too long to translate in one go. Solution:

Always use section-by-section approach (see Step-by-Step Process above)
If a single section is too long, split into subsections:
- "Introduction"을 3.1, 3.2, 3.3으로 나눠서 번역
Create separate output files for each section if needed
User can combine sections later

Recommended approach for very long papers:

논문 구조:
- Abstract (짧음 - 한번에 번역)
- Introduction (중간 - 한번에 또는 2개 파트로)
- Related Work (긺 - 3-4개 파트로 나눔)
- Methods (매우 긺 - 5-6개 파트로 나눔)
...

각 파트를 순차적으로 번역하되, 사용자 승인 없이 자동으로 진행

Formatting Preservation Checklist

OCR artifacts removed (page numbers, headers, footers)
Line breaks normalized (meaningless breaks removed)
Hyphenated words merged
All headers maintained with same level
Lists formatted correctly
Code blocks preserved with syntax highlighting
Tables aligned properly
Links functional
Citations in original format
Equations/formulas unchanged
Image references intact
YAML header from header.yaml prepended to first section
YAML header contains embed-resources: true for image embedding
Translated content saved to {filename}_ko.md
Each section appended to the same output file
No duplicate YAML headers in subsequent sections
References/Bibliography section: Header translated, content kept in English
Appendix section: Header translated, content kept in English

paper-translator-korean

Install Skill

SKILL.md