thesis-reviewer

Thesis Review and Semantic Review

Academic Integrity and the Legitimate Use of "Humanization/Style Review"See the root README.md's"Design Boundaries and Academic Integrity". For the Part B cross-chapter topic list, see references/review-checklist.md(supports custom entries + replaceable examples).

Review Modes

Select the execution mode according to user needs:

Mode	Trigger Scenario	Execution Content
Comprehensive Review	"blind review" "review" "comprehensive check"	Part A (macro review) + Part B (semantic close review) + Part C (AI trace review, performed by default)
Macro Review	"review comments" "thesis quality" "can graduate or not"	Execute Part A only
Semantic Close Review	"semantic repetition" "terminology consistency" "redundancy check" "cross-chapter comparison"	Execute Part B only
AI Style Review	"AI style" "AI traces" "humanization check" "writing style review" "remove AI flavor"	Execute Part C only (may also be included by default in comprehensive review/single-chapter review)
Single-Chapter Review	"review Chapter N" "check this section"	Part B first round for the specified chapter + Part C (AI trace review, performed by default), as well as Part A dimensions 4-5

---

Part A: Macro Review (Supervisor/Blind Review Perspective)

Review Dimensions

A1 Topic Selection and Literature Review

• Value of Topic Selection: whether it has clear theoretical significance or engineering application value, and whether it is closely tied to disciplinary frontiers or national/industry needs
• Literature Review Quality: whether the literature investigation is comprehensive and objective; whether citations are authoritative and highly current
• Problem Orientation: whether it accurately identifies deficiencies based on the literature review and naturally leads to the scientific question

A2 Theoretical Foundation and Research Methods

• Conceptual Accuracy: whether basic concepts are clearly defined and the theoretical foundation is solid
• Methodological Soundness: whether models, algorithms, and experimental methods are scientific and reasonable
• Technical Route: whether the overall technical route is feasible and the logical chain is rigorous

A3 Workload and Research Results

• Workload Sufficiency: whether the amount of experimental data, scale of system development, and depth of case analysis meet the standard
• Argumentation and Analysis: whether it goes deep into the mechanism level rather than merely describing phenomena
• Innovation and Contribution: whether there is innovation at the theoretical/methodological/application level, and whether it is convincing

A4 Thesis Logic and Structure

• Chapter Structure: whether the chapter arrangement is reasonable (introduction→theory→modeling→validation→conclusion), and whether the logical connections between chapters are strong
• Paragraph coherence: Whether transitions within and between paragraphs are natural, and whether there are logical jumps, argumentative gaps, or contradictions

A5 Academic Writing and Formatting Standards

• Language expression: Whether it is rigorous, concise, and objective, and whether there is colloquialism or subjective speculation
• Figure and table quality: Whether clarity, labels, units, and figure/table captions are self-explanatory
• Typesetting standards: Equation numbering, citation format, and consistency of terminology

Macro-level review output

1. Overall evaluation: Greatest highlight + most fatal flaw + whether it reaches degree level
2. Dimension-specific revision suggestions: Point out problems and explain the reasons
3. Actionable suggestions: Specific guidance on how to supplement data/comparative experiments/redraw figures/restructure logic

---

Part B: Semantic close review (chapter by chapter + cross-chapter + terminology)

Problem classification labels

Label	Category	Definition
REP	Semantic repetition	The same fact/conclusion is repeatedly stated in different locations
RED	Semantic redundancy	Low information content within a paragraph, tautology, filler modifiers
CON	Statement conflict	The same concept is expressed in incompatible ways in different locations
TRM	Terminology inconsistency	The same variable/abbreviation/term name is not unified
LNK	Transition break	Chapter-opening/chapter-ending transitions are disconnected from the content of adjacent chapters
FMT	Formatting flaw	Typesetting issues such as equation environments, citation format, and figure/table numbering

Three-round review process

First round: Intra-chapter review

Read all .tex files chapter by chapter, and produce a problem list for each chapter.

Execution order: Core technical chapters first (Chapters 3→4→5→6→2→1→7).

Check items：

1. Scan paragraph by paragraph for intra-paragraph redundancy (tautology, information-free modifiers)
2. Inter-section repetition (different sections in the same chapter repeatedly describe the same content)
3. Consistency of terminology and symbols within the chapter
4. Transitional connections at the beginning and end of each section
5. Depth of argumentation (whether it only describes phenomena without mechanism analysis)

Second round: Cross-chapter review

For high-frequency repetition/conflict domains, read the relevant sections of two chapters simultaneously for comparison.

General cross-chapter review topics：

• Whether core concept definitions are repeatedly explained from scratch in multiple chapters
• Whether algorithm/method descriptions are written once each in the review, main text, and case study
• Whether the problem definition (mathematical model) is repeated in the modeling chapter and the case study chapter
• Whether process descriptions are repeatedly narrated in multiple places
• Whether the study area/experimental background is repeated in the data chapter and the case study chapter
• Whether parameter values/names are consistent in the modeling chapter, simulation chapter, and appendix
• Whether the definitions of comparative concepts such as "baseline method" are consistent throughout the full text

For project-specific cross-chapter review topics, see review-checklist.md。

Third round: Terminology and symbol consistency

Establish a terminology-symbol comparison table and check the full text:

1. Consistency of Chinese expressions for core terminology
2. Consistency of mathematical symbol notation (e.g., $v_{\max}$ vs $v_{\text{max}}$）
3. English abbreviations: full name on first occurrence + unified abbreviation thereafter
4. Validity of figure/table numbering references (\ref Does it point to a deleted figure/chart)
5. Dimensions and units are consistent throughout the text (m/s vs km/h, etc.)

Auxiliary operation: rapid keyword scanning

Before detailed review, you can first use keyword search to locate high-frequency repeated areas:

• For each suspicious term variant, search the entire .tex for the number of occurrences and locations
• Mark terms that appear in more than 3 files as "cross-chapter repetition candidates"

---

Part C: AI style review

Consistent with the humanizer-zh rules of academic-writing; for details of the rules, see ../academic-writing/references/humanizer-zh-patterns.md. When executing Part C, read that reference and use the pattern list and check items in it.comprehensive reviewandsingle-chapter reviewwhenexecute by default Part C, without the user separately requesting "AI style"; the user may also request only "AI style review only" to run Part C separately. It aims to identify the stylistic features typical of AI-assisted writing, so as to facilitate targeted rewriting and improve naturalness and academic human touch,not used to determine whether plagiarism has occurred, and is only for reference in writing quality and style optimization.

Check dimensions

Dimension	Description	Typical manifestation
Mechanical transitions	Transition words/conjunctions are too concentrated and predictable	Strings such as "first, second, in addition, however, in summary" appear; frequent paragraph openings such as "in view of this," "on this basis," "from the perspective of..."
Uniform sentence length	Sentence length lacks rhythmic variation	Most sentences are similar in length, lacking short punchy sentences or long elaborative sentences
Abstract boilerplate	All-purpose sentence patterns with low information content	Vague expressions such as "is of great significance," "is worthy of attention," "in terms of...," "in a certain sense," "it is worth noting that"
AI habitual words	Buzzwords and boilerplate words commonly found in AI-generated text	Such as "closed loop," "empowerment," "grip point," "implementation," "pain point," "connect through," "iteration," "precipitation," "dimension," "scenario," "link," "underlying logic," "end-to-end," "deep integration," "organic combination," "effective/significant improvement," "systematicness," "holism," "synergy," "coupling," "paradigm," "reproducible," "scalable," etc.; mark them if they do not conform to disciplinary conventions or appear vague
Habitual sentence patterns and structures	Fixed collocation sentence patterns and "A-B-C" parallel structures	"First... second... finally..."; "one is... two is... three is..."; repeated use of "on the basis of/in the context of/under the framework of..."; "the complete link/closed loop from... to..."; "realizes the organic combination/deep integration of... and..."; "provides.../lays... for..."; each paragraph begins with "this article" or "this study"; excessive use of "can," "be able to," "can effectively," etc.
Monotonous vocabulary	The same concept is repeatedly expressed with the same word, lacking synonymous substitution	The same verb/noun is repeated multiple times within the same paragraph; vocabulary diversity outside terminology is low
Overly neat parallel structures	Deliberately symmetrical sentence patterns	"On the one hand... on the other hand" appears in pairs; consecutive sentences with "..., thereby..." "..., and then..."
Identical paragraph structures	The structure within paragraphs is too regular	Each paragraph follows a fixed pattern of "general-specific-general" or "phenomenon-cause-conclusion"

Examples of AI habitual words and habitual sentence patterns (for reference during review):

• Habitual words: closed loop, empowerment, lever, implementation, pain point, connect through, iteration, consolidation, dimension, scenario, link, underlying logic, end-to-end, deep integration, organic combination, effective improvement, significant improvement, systematic, holistic, synergistic, coupling, paradigm, reproducible, scalable, all-round, one-stop, granularity (in non-professional contexts).
• Habitual sentence patterns/structures: first-second-finally / one-two-three; on the basis of..., in the context of..., under the framework of...; the complete... from... to...; achieved the... of... and...; provided... for..., laid the foundation for...; repeated "this paper..." "this study..." at the beginning of paragraphs; consecutive "can..." "is able to..." "can effectively...".

Issue markers

Marker	Category	Definition
AIT	Typical AI transition	Mechanical transition words are concentrated and can be replaced with more natural connections
AIS	Typical AI sentence pattern	Abstract boilerplate, all-purpose sentence patterns; recommend changing to specific expressions
AIV	AI habitual words	Common AI buzzwords/boilerplate words appear (such as closed loop, lever, implementation, etc.); recommend replacing or deleting according to disciplinary conventions
AIF	Habitual sentence patterns/structures	A-B-C-style parallel structures, fixed collocation sentence patterns (such as consecutive use of "on the basis of...", "lay... for..."); recommend breaking down or rewriting
AIW	Monotonous vocabulary	Too much repeated wording within the same paragraph; recommend synonym replacement or omission
AIP	Similar structures	Paragraph or sentence group structures are too neat and formulaic

Execution method

1. Chapter-by-chapter scan: In the order of chapters 3→4→5→6→2→1→7, perform a paragraph-level scan of each chapter's main text (excluding figures, tables, and formulas).
2. Compare against humanizer-zh: Combine with ../academic-writing/references/humanizer-zh-patterns.md the content/language/style/communication modes and quick checklist to identify hits.
3. Mark locations: For suspected places, mark file:line number (or section number), type (AIT/AIS/AIV/AIF/AIW/AIP, or the corresponding humanizer-zh pattern), original text excerpt, and brief revision suggestion.
4. Optional tools: If the project contains academic-writing skill's scripts/ai_detector.py、scripts/text_analyzer.py, you may run detection on exported paragraphs (such as sentence length distribution, transition word density) to help locate high-risk paragraphs; use the results together with manual review.

Output format

AI style review output can be merged into the Part B issue list table,Typecolumn uses AIT/AIS/AIV/AIF/AIW/AIP;Suggested revisioncolumn gives specific rewriting directions (such as "change to specific conclusions/data", "shorten sentence length", "replace transition words", "replace habitual words", "break down A-B-C structure"). You may addAI trace diagnosissummary: list typical paragraphs that hit humanizer-zh patterns and corresponding suggestions. You may also output a separate "AI style review checklist", with the same format as the issue list table.

---

Output format

Issue list table (produced in each round of Part B; Part C may be merged in or made into a separate table)

| 编号 | 类型 | 位置（文件:行号） | 问题描述 | 建议修改 | 优先级 |
|------|------|------------------|---------|---------|--------|
| R01  | REP  | 3.2.0:45 & 4.2.0:12 | 两处完整解释坡度阻力公式 | 第4章仅引用第3章 | 高 |
| AI01 | AIT  | 4.2.0:12 | 段首连续「首先」「其次」 | 改为自然衔接或合并句式 | 中 |

Type: Part B uses REP/RED/CON/TRM/LNK/FMT; Part C uses AIT/AIS/AIV/AIF/AIW/AIP.

Priority: High (affects academic rigor/reviewers will point it out directly), Medium (affects reading experience), Low (minor detail flaws).

Comprehensive Review Report (produced when Part A + B are aggregated)

1. Overall Evaluation and Scoring (by the five dimensions A1-A5)
2. Summary of Key Issues (extract high-priority items from the Part B issue list)
3. AI Trace Diagnosis(Part C): If Part C has been executed, summarize typical paragraphs matching humanizer-zh patterns and revision suggestions
4. Revision Priority Ranking and Actionable Suggestions

---

Execution Workflow

1. Global Overview: Quickly read through the abstract, introduction, and conclusion to grasp the research objectives and core conclusions
2. Macro Review(Part A): Cross-check by the five dimensions A1-A5 (executed during comprehensive review)
3. Semantic Scrutiny(Part B): Execute step by step according to the three-round process (executed during comprehensive review or semantic scrutiny)
4. AI Style Review(Part C): Scan chapter by chapter according to humanizer-zh rules and inspection dimensions;executed by default during comprehensive review and single-chapter review, and Part C alone may also be executed upon user request
5. Aggregated Output: Merge all issue lists, sort by priority, and output the complete review report (including a summary of AI trace diagnosis, if Part C has been executed)