academic-thesis-writer-vi

name	academic-thesis-writer-vi
description	Write master's-level academic theses in Vietnamese with formal scholarly style, IEEE citations, and bilingual terminology. Use when writing thesis chapters, research papers, or academic content in Vietnamese.

Vietnamese Academic Thesis Writer (Master's Level)

Overview

This skill guides writing chapters of master's-level essays, theses, and dissertations in Vietnamese following a professional, formal academic style with properly formatted citations.

When to Use

Activate this skill when the user requests Vietnamese academic content at master's level, including but not limited to:

• Writing a specific chapter (e.g., "write Chapter 2 theoretical background")
• Writing a section/subsection within a chapter (e.g., "write section 3.2 system architecture")
• Writing custom content on request (e.g., "write a comparison of TF-IDF and BM25")
• Editing / improving existing academic content
• Creating or editing a references list
• Writing research reports, graduation projects, or course term papers (tiểu luận môn học) to academic standards

Vietnamese-language signals that indicate this skill applies: tiểu luận, luận văn, thạc sĩ, khóa luận, chương, tổng quan, cơ sở lý thuyết, phương pháp nghiên cứu, báo cáo nghiên cứu, đồ án tốt nghiệp, viết chương, viết mục, viết phần. English equivalents: academic, thesis, dissertation, master's paper.

Operating Principles

IMPORTANT: This skill writes exactly what the user requests. Do NOT write the entire document or add sections the user did not ask for.

Workflow:

1. Identify the request: The user specifies which chapter, section, or content to write.
2. Read citation.md (critical, do this FIRST): Also read any .bib companion with the same basename. These are the single source of truth for references. From them:
   • Find the last citation number [M]. New references start from [M+1].
   • Note existing entries so you can reuse their [N] instead of creating duplicates.
3. Distinguish source materials from report content (see Rule 7): Any file or note provided as author-side guidance (outlines with annotations, rubrics, task prompts, review instructions, internal specifications) is for orientation only. It MUST NOT appear in the final report text.
4. Scan context of existing chapters: If previous chapters exist, read them to:
   • Note terms already introduced bilingually. Do not re-introduce them, use the abbreviation.
   • Ensure consistent tone and style with previous chapters.
5. Gather technical context: Read PRDs, technical docs, source code (if available) to understand the content accurately.
6. Write only the requested section: Apply all rules below. Do NOT add a per-chapter references list. All references go to citation.md.
7. Update citation.md (and .bib if present): Before appending, check whether each reference is already in the ledger; if so, reuse its [N]. Append only genuinely new references with [M+1], [M+2], …, and mirror the appends into the .bib companion if present (see Rule 3, subsections 5 and 6). Update the mapping table if one exists.
8. Review: Run the checklist before returning the result.

MANDATORY RULES

Rule 1: Language and Tone

Strict compliance required:

1. Point of view: Use third person or passive voice. NEVER use first person ("tôi", "chúng tôi", "em", "nhóm tác giả"). Instead use: "đề tài", "hệ thống", "nghiên cứu này", "tác giả".

2. Formal tone: Use scientific, precise language. Avoid colloquial, informal, or emotionally expressive words.

3. Bilingual Vietnamese-English for technical terms: When mentioning a technical term for the first time, ALWAYS write it in Vietnamese followed by the English equivalent in parentheses, plus the abbreviation separated by a comma if applicable.

Standard patterns (use comma as separator, NEVER em dash):

Truy hồi thông tin (Information Retrieval, IR)
Xử lý Ngôn ngữ Tự nhiên (Natural Language Processing, NLP)
Mô hình Không gian Vector (Vector Space Model, VSM)
Tần suất thuật ngữ - Tần suất nghịch đảo tài liệu (Term Frequency-Inverse Document Frequency, TF-IDF)
Độ tương đồng cosine (Cosine Similarity)
Tách từ (word segmentation)
Từ dừng (stopwords)

For compound terms (e.g. TF-IDF), use a hyphen between the two English sub-terms inside the parentheses.

4. Subsequent mentions: May use the abbreviation introduced at first occurrence.

Correct example:

Truy hồi thông tin (Information Retrieval, IR) là lĩnh vực nghiên cứu...
Trong phạm vi đề tài, hệ thống IR được thiết kế dựa trên...

Incorrect examples:

❌ "Tôi sẽ xây dựng hệ thống tìm kiếm..."
❌ "Chúng ta cần phải sử dụng TF-IDF..." (not introduced yet)
❌ "Cái này rất quan trọng vì..."
❌ "Truy hồi thông tin (Information Retrieval — IR)..." (em dash, forbidden by Rule 8)

---

Rule 2: Chapter Structure

Each chapter MUST follow a hierarchical numbering system:

# CHƯƠNG N. CHAPTER TITLE (UPPERCASE)

## N.1. Level 1 heading

### N.1.1. Level 2 heading

#### N.1.1.1. Level 3 heading (if needed)

Numbering rules:

• Chapter: CHƯƠNG 1, CHƯƠNG 2, ...
• Level 1: 1.1., 1.2., 1.3., ...
• Level 2: 1.1.1., 1.1.2., ...
• Level 3: 1.1.1.1. (limit usage, only when truly necessary)

Heading separators: if a heading needs a subtitle connector (e.g. "Bước 1: Tiền xử lý"), use a colon (:) or restructure into plain prose. NEVER use em dash in headings, table labels, or figure captions.

Chapter titles: ALL UPPERCASE. Section titles: Capitalize first letter, rest in lowercase.

---

Rule 3: Citations and References

Use IEEE citation style (numbers in square brackets):

1. In-text: Number as [1], [2], [3], ... in order of first appearance.

2. Citation placement: Immediately after the cited content, before the period.

In-text citation patterns:

Theo nghiên cứu của Manning và cộng sự [2], bài toán truy hồi thông tin...
Mô hình Vector Space Model được đề xuất bởi Salton (1975) [3].
Tách từ tiếng Việt là bước tiền xử lý bắt buộc [7].

3. Centralized references file (citation.md): ALL references are stored in a single file (e.g., docs/report/citation.md), NOT at the end of each chapter. This is the single source of truth for citation numbering across the entire document.

   Why centralized? Scanning multiple chapter files for the last citation number is error-prone and slow. A single file makes it easy to:

   • Check the last number [M] instantly
   • Reuse existing references without duplicating
   • Maintain consistency across all chapters

   Format (IEEE style, one blank line between entries):

   [1] Author_Name, "Paper title," _Journal/Conference Name_, vol. X, no. Y, pp. ZZ–ZZ, Year.
   
   [2] Author_Name, _Book Title_, Publisher, Year.
   
   [3] Author_Name, "Report title," _Organization Name_, Year.
   

   Formatting rules for citation.md:

   • Each [N] entry on its own line, separated from the next by one blank line. This ensures each reference renders as its own paragraph when the document is exported to DOCX or PDF (e.g., via Pandoc). Entries packed together without blank lines will collapse into a single paragraph in Word.
   • One blank line after the heading # TÀI LIỆU THAM KHẢO and before the optional mapping table.
   • The last citation number [M] remains easy to scan because entries are ordered sequentially; a blank-line separator does not impede lookup.

   Optional mapping table: citation.md may include a --- separator followed by a table mapping each [N] to the chapters that cite it. Update this table when adding new references or writing new chapters.

4. IMPORTANT: Citation number [N] in the chapter text MUST match EXACTLY with [N] in citation.md. Review everything for consistency when finished.

5. CROSS-CHAPTER CONTINUITY (most important rule):

   Mandatory procedure:

   • Step 1: Read citation.md to find the last citation number [M].
   • Step 2: Before adding a reference, check whether it already exists in citation.md (match by DOI, or by first author + year + title). If yes, reuse that [N] — do NOT create a new entry.
   • Step 3: When the reference is genuinely new, assign [M+1], [M+2], etc. and append them to citation.md after writing the chapter.

   Example:

   citation.md has [1] through [10]
   → Chapter 4 reuses [4], [5], [9] and adds new [11], [12]
   → After writing, append [11] and [12] to citation.md
   → citation.md now has [1] through [12]
   

   If citation.md does not exist (e.g., first chapter), create it and start from [1].

   Chapter files should NOT contain a "Tài liệu tham khảo" section at the end. All references live in citation.md only.

6. Optional BibTeX companion file for reference managers: If a .bib file exists in the same directory with the same basename as citation.md (e.g., citation.md + citation.bib), treat the two files as a coupled pair and keep them in sync. This is the single feature that lets the author import the thesis bibliography into Mendeley, Zotero, EndNote, or any LaTeX pipeline.

   Activation rule: The .bib file is opt-in. Do NOT create one if the author has not already placed it in the directory. When it exists, every update to citation.md requires a matching update to the .bib; when it is absent, ignore this subsection entirely.

   Entry type: Choose based on the source:

   • @article{...} — journal articles
   • @inproceedings{...} — conference proceedings
   • @book{...} — books
   • @techreport{...} — institutional reports (IDC, CMU tech reports, etc.)
   • @misc{...} — anything else (online resources, standards)

   Citation key convention: firstAuthorLastName + year + firstSignificantTitleWord, all lowercase (e.g., mihalcea2004textrank, vaswani2017attention). Stable keys keep Mendeley imports idempotent and let co-authors cross-reference.

   Link fields (preferred order): Include at least one of the following so the reference becomes clickable after import:

   • doi = {10.xxxx/yyy} — preferred; Mendeley/Zotero resolve this to the publisher page automatically.
   • url = {https://...} — use when the work has no DOI (ACL Anthology papers, JMLR, preprints hosted on personal pages, IDC landing pages).
   • eprint = {arxiv-id} + archivePrefix = {arXiv} — use for arXiv preprints; add a mirrored url = {https://arxiv.org/abs/<id>} for convenience.
   • isbn = {...} — for books.

   Sync procedure: When appending [N] to citation.md, extend the subsection 5 check to the .bib too (look up the proposed key and DOI); if already present, reuse and skip the append. Only for genuinely new entries, append a BibTeX block to the .bib, separated from the previous entry by one blank line (same rationale as Rule 3 subsection 3: DOCX/PDF export via Pandoc treats blank-separated blocks as paragraphs). Keep entry ordering in .bib aligned with [N] in citation.md so the two files are easy to diff.

   Filling missing links automatically: A helper script is bundled with this skill for enriching entries that lack doi/url/eprint:

   python3 scripts/fetch_citation_links.py path/to/citation.bib
   

   The script queries Crossref first (then arXiv as fallback), gates each match on title-word overlap to avoid false hits, and edits the .bib in place. Use --dry-run to preview changes, --min-overlap to tune the matching threshold (default 0.7), or --min-score to adjust the Crossref score floor. Run it after adding a batch of new entries, then inspect the log — any fail: lines need manual follow-up via Google Scholar or the publisher site.

---

Rule 4: Paragraphs and Argumentation

1. Minimum 3 to 5 sentences per paragraph. Each paragraph should focus on ONE main idea.

2. Paragraph argumentation structure:

   • Topic sentence → Supporting evidence or elaboration → Concluding or transition sentence

3. Paragraph linking: Use formal transition words/phrases:

- Opening: "Trong bối cảnh...", "Với sự phát triển của..."
- Addition: "Bên cạnh đó,", "Ngoài ra,", "Hơn nữa,"
- Contrast: "Tuy nhiên,", "Mặc dù vậy,", "Ngược lại,"
- Result: "Do đó,", "Vì vậy,", "Từ đó,"
- Enumeration: "Thứ nhất,", "Thứ hai,", "Cuối cùng,"
- Example: "Cụ thể,", "Chẳng hạn,", "Ví dụ,"

4. When listing multiple points: Use the pattern "Thứ nhất, ... Thứ hai, ... Thứ ba, ..." with each part starting in bold, followed by a detailed explanation.

---

Rule 5: Tables and Mathematical Notation

1. Tables: Use Markdown tables for comparisons, technical specifications.

2. Mathematical notation: Use Unicode characters when writing in Markdown:

Set notation: D = {d₁, d₂, ..., dₙ}
Subset: R ⊆ D
Calculation: w(t,d) = tf(t,d) × log(N/df(t))

3. Complex formulas: Present on separate lines, clearly explain each symbol.

---

Rule 6: Scope and Limitations

When writing the scope and limitations section, you MUST:

1. Categorize clearly by aspects: language, data, algorithms, architecture, evaluation, security.

2. Scope: State clearly WHAT is in scope, using bullet points.

3. Limitations: Present honestly WHAT is NOT covered and why. This demonstrates academic self-awareness. Limitations are not a weakness but scientific honesty.

---

Rule 7: Source materials vs report content

When writing an academic report, distinguish between two kinds of content:

• Report content: what appears in the final document (prose, tables, figures, references).
• Source materials: context provided to the author (outlines with annotations, rubrics, task prompts, review instructions, internal specifications, meta-comments, drafts).

Source materials are for orientation only. They MUST NOT be reproduced, paraphrased, or structurally mapped into the final report text.

Typical source materials to keep out of the report:

• Grading rubrics, scoring guides, evaluation criteria (from any source: school, journal reviewer, supervisor, client).
• Outline annotations, especially markers that tie sections to external criteria (e.g. <!-- ✅ Criterion 3 -->, /* TODO */, "→ đáp ứng tiêu chí X").
• Instructor, reviewer, or client prompts used during authoring (e.g. "the user asked to write...", "as per the brief...").
• Institutional meta-context the report is written for but not about (e.g. course codes, cohort names, supervisor preferences).
• Deadlines, project-management notes, or other author workflow artefacts.
• Any table that maps report sections to external scoring criteria.

Why: A report must read as a self-contained academic artefact. Leaking source materials signals "written to satisfy an external form" and undermines scholarly tone.

How to apply:

• Frame design principles as the project's own: "nguyên tắc của đề tài", "mục tiêu của nghiên cứu"; NOT "theo yêu cầu của [X]", "để đạt [tiêu chí]".
• Any section that summarizes results (conclusion, abstract, summary) presents findings qualitatively and quantitatively on their own terms; it does NOT reproduce rubric or criteria.
• If source-material language appears in user requests or drafts, rewrite it in terms of the project's own rationale.

---

Rule 8: No em dash anywhere

The em dash character (—, Unicode U+2014) is NOT used in this skill's output. This rule supersedes any earlier habit of using em dash for bilingual term abbreviations, clause insertions, or heading subtitles.

Replacements:

Former em dash use	Correct replacement
Bilingual abbreviation: (Natural Language Processing — NLP)	Comma: (Natural Language Processing, NLP)
Clause insertion: "dữ liệu — đặc biệt văn bản — tăng"	Comma or parentheses: "dữ liệu (đặc biệt văn bản) tăng"
Heading subtitle: "Bước 1 — Tiền xử lý"	Colon: "Bước 1: Tiền xử lý"
Table or figure caption: "Hình 2.1 — Pipeline"	Colon: "Hình 2.1: Pipeline"
Connective: "tính chất này — đảm bảo..."	Comma plus conjunction: "tính chất này, đảm bảo..."

Allowed dash variants:

• En dash (–, U+2013) for numeric ranges: 2001–2007, pp. 404–411.
• Hyphen (-, U+002D) for compound terms inside parentheses: (Term Frequency-Inverse Document Frequency, TF-IDF).

Verification: Before returning the chapter, grep the output for the — character. If any instance is found, replace it per the table above.

---

STANDARD CHAPTER STRUCTURE

Chapter 1: OVERVIEW

1.1. Introduction
  1.1.1. Research background (why is this topic important?)
  1.1.2. Problem statement (what is the specific problem?)
  1.1.3. Challenges (domain-specific difficulties?)
  1.1.4. Modern trends and the topic's position
1.2. Objectives
  1.2.1. General objective (1 concise sentence)
  1.2.2. Specific objectives (3-5 sub-objectives, numbered)
1.3. Scope and limitations
  1.3.1. Research scope
  1.3.2. Limitations

Chapter 2: THEORETICAL BACKGROUND

2.1. Foundational concepts (formal definitions)
2.2. Related models / algorithms
  2.2.1. Model A (theory + formulas)
  2.2.2. Model B
2.3. Technologies used
  2.3.1. Framework / Library A
  2.3.2. Framework / Library B
2.4. Related work

Chapter 3: METHODOLOGY / SYSTEM DESIGN

3.1. Overall research methodology
3.2. System architecture (overview diagram)
3.3. Detailed design
  3.3.1. Module A
  3.3.2. Module B
3.4. Data processing flow
3.5. API / Interface design

Chapter 4: EXPERIMENTS AND RESULTS

4.1. Experimental environment
4.2. Dataset
4.3. Evaluation methods
4.4. Experimental results
  4.4.1. Quantitative results (tables, charts)
  4.4.2. Qualitative results (analysis, observations)
4.5. Comparison and discussion

Chapter 5: CONCLUSIONS AND FUTURE WORK

5.1. Conclusions
5.2. Contributions
5.3. Limitations
5.4. Future work

Note: The 5-chapter structure above is a common template. Adapt chapter count and titles to the document type (thesis, term paper, journal article, technical report). Regardless of structure, Rule 7 applies to every section.

---

REFERENCE EXAMPLE

Below is a sample demonstrating the correct writing style for a paragraph in the Overview chapter. Use this as a style reference.

### 1.1.1. Bối cảnh nghiên cứu

Trong bối cảnh bùng nổ thông tin số hiện nay, lượng dữ liệu phi cấu trúc
(đặc biệt là văn bản) đang tăng trưởng với tốc độ chưa từng có. Theo ước
tính của International Data Corporation (IDC), tổng lượng dữ liệu toàn cầu
đã đạt hơn 120 Zettabyte vào năm 2023 và được dự báo sẽ tiếp tục tăng gấp
đôi sau mỗi hai năm [1]. Trong đó, dữ liệu văn bản chiếm một tỷ trọng đáng
kể, bao gồm các bài báo trực tuyến, bài đăng trên mạng xã hội, blog, tài
liệu kỹ thuật và nhiều nguồn nội dung số khác.

Truy hồi thông tin (Information Retrieval, IR) là lĩnh vực nghiên cứu về
các phương pháp tìm kiếm, trích xuất và xếp hạng các tài liệu có liên quan
đến nhu cầu thông tin của người dùng từ một tập hợp tài liệu lớn [2].

Note the key characteristics of this sample:

• Starts with broad context, narrows down gradually
• Specific data with citation [1]
• Bilingual terminology with comma separator: "Truy hồi thông tin (Information Retrieval, IR)"
• Parenthetical insertion uses parentheses, not em dash: "dữ liệu phi cấu trúc (đặc biệt là văn bản)"
• Scientific, formal tone, no personal emotion
• Long but coherent sentences, not convoluted

---

CHECKLIST BEFORE COMPLETION

After writing each chapter, REVIEW against this checklist:

• Citation numbers [N] in text match entries in citation.md
• New references have been appended to citation.md (with correct [M+1] numbering)
• No per-chapter references section: chapter file does NOT contain "Tài liệu tham khảo" at the end
• No duplicate references: each new reference was checked against citation.md (and .bib if present) before appending; matches reuse the existing [N] instead of getting a new one.
• Technical terms introduced with English equivalent and abbreviation at first occurrence, using a comma separator (not em dash)
• No first person (tôi, chúng tôi, em, nhóm tác giả)
• Section numbering follows hierarchical system (N.N.N.)
• Each paragraph has at least 3 sentences, focused on 1 main idea
• Tables have clear titles
• No Vietnamese spelling errors (especially dấu hỏi/ngã tone marks)
• Tone consistent and formal throughout
• Scope/limitations section presented honestly, no hidden weaknesses
• Transition sentences between paragraphs and sections flow smoothly
• No em dash (—) anywhere in the output (headings, prose, captions, tables). Grep for the — character to verify (see Rule 8).
• No source-material language or mapping: the report stands on its own without referencing rubrics, task prompts, external evaluation criteria, or institutional meta-context (see Rule 7).
• BibTeX sync (only when .bib exists): every [N] added to citation.md has a corresponding entry in the .bib with a stable key and at least one link field (doi, url, eprint, or isbn). If some entries lack links, note them for follow-up via scripts/fetch_citation_links.py (see Rule 3, subsection 6).

---

COMMON MISTAKES TO AVOID

#	Mistake	Incorrect Example	Correction
1	First person usage	"Tôi sẽ xây dựng hệ thống..."	"Đề tài xây dựng hệ thống..."
2	Term without English equivalent	"Mô hình không gian vector"	"Mô hình Không gian Vector (Vector Space Model, VSM)"
3	Wrong citation number	[5] in text ≠ [5] in references	Review all citations end to end
4	Single-sentence paragraph	"TF-IDF là sơ đồ trọng số quan trọng."	Expand with 2-4 more sentences of explanation/evidence
5	Abbreviation without intro	"Hệ thống IR sử dụng VSM" (not explained)	Introduce fully at first occurrence
6	List without explanation	"Ưu điểm: nhanh, chính xác, dễ dùng"	Develop each point into a short paragraph
7	Vague conclusion	"Hệ thống hoạt động tốt"	State specific metrics: "Response time under 500ms"
8	Em dash for bilingual or clause	"TextRank (TextRank Algorithm — TRA)"	Use comma: "TextRank (TextRank Algorithm, TRA)"
9	Em dash in heading	"### 2.2.1. Bước 1 — Tiền xử lý"	Use colon: "### 2.2.1. Bước 1: Tiền xử lý"
10	Source-material mapping in report	"Bảng ánh xạ các mục của báo cáo với tiêu chí đánh giá..."	Remove. The report summarizes findings on its own terms, not against an external checklist.
11	External-context leakage	"...nhằm đáp ứng yêu cầu của [giảng viên/reviewer/khách hàng]"	"...nhằm đáp ứng mục tiêu của đề tài."

---

FINAL NOTES

• This skill applies to master's-level essays, theses, and dissertations at Vietnamese universities, as well as course term papers (tiểu luận môn học), journal articles, and technical reports that follow the same academic conventions.
• The writing style is extracted from real samples that received high quality ratings.
• Can be combined with other skills (e.g., api-documentation, architecture) for detailed technical chapters.