Logic Module Reference
Purpose: Check logical coherence, introduction funnel, heading lead-ins, literature review quality, chapter mainline, and cross-section closure.
For chapter-level rewrite planning, also read ../writing/thesis-writing-guide.md. Keep logic as the diagnostic route, but use the guide to turn findings into a thesis-specific mainline plan.
AXES Model (Paragraph-Level Coherence)
| Component | Role | Example |
|---|---|---|
| Assertion | Clear topic sentence | "注意力机制能够提升序列建模效果。" |
| Xample | Supporting evidence/data | "实验中,注意力机制达到95%准确率。" |
| Explanation | Why evidence supports claim | "这一提升源于其捕获长程依赖的能力。" |
| Significance | Connection to broader argument | "这一发现为本文架构设计提供了依据。" |
Heading Lead-In Check (S1)
Rule: Every chapter, section, subsection, and content-bearing subsubsection must have a lead-in paragraph before any list, figure, table, formula, or child heading.
Lead-in minimum: State what will be discussed, why here, connection to previous content, and preview of internal structure.
Detection: Script scans \chapter, \section, \subsection, \subsubsection, \paragraph — flags if first child is non-prose content.
Chapter Intro Specialization (承上启下)
S1 只判断"有没有导语"。对正文各章(第 2 章至结论前、且含下级小节)的章引言,脚本另做承上启下专项检查(% 章引言 ... [Script]),与 S1 互补:
- 承上缺失 / 启下缺失(Major/P1):章引言未承接前章(无章节号/桥接),或未交代本章问题与各节安排。
- 相对指代(Minor/P2):出现"上一章/上文",建议改用章节号"第 X 章"。
- 篇幅过简 / 过长(Minor/P2):偏离"1~2 段、约 300~500 字"的约定。
绪论(第 1 章)由 _check_introduction_funnel 负责,章引言检查按标题显式排除,零重叠。改写指导见 ../writing/thesis-writing-guide.md 的"正文章引言"一节。
Literature Review Quality (A1-A4)
| Check | Rule | Detection |
|---|---|---|
| A1: Topic clustering | Organize by theme, not author/year listing | Script: regex for 3+ consecutive "Author(Year) proposed..." |
| A2: Critical analysis | Each topic group needs evaluative commentary | LLM judgment required |
| A3: Gap derivation | Last paragraph must identify research gap | Script: keyword scan in final 10 lines |
| A4: Funnel citation density | Citations should narrow from broad to specific | LLM judgment required |
Cross-Section Closure (C3)
Rule: Contribution claims in introduction must be echoed in conclusion.
Detection: Script extracts contribution keywords from introduction, checks for response keywords ("验证了", "证明了", "实验表明") in conclusion. Missing echo → Major/P1.
Intro Mainline Checks (--intro-mainline)
uv run python -B scripts/analyze_logic.py thesis.tex --intro-mainline绪论主线四项专项检查,全部 [Script] 启发式,仅在传入该 flag 时运行(默认行为不变):
| Check | Rule | Severity |
|---|---|---|
| L-SCI | 科学问题(表格“科学问题”列或枚举条目)不得是短名词短语,须含对象-问题-方法三要素 | Major/P1 |
| L-MAP | 科学问题/研究内容/创新点条数应闭合;正文声明不等量(如“工程验证贡献,不与……等量”)则降级 Info | Major/P1 |
| L-FUN | 绪论首段须完成 领域背景 -> 技术瓶颈 -> 本文 三层漏斗 | Minor/P2 |
| L-DOM | 标题写“国内外研究现状”就必须分述国内/国外,或声明按主题混排 | Info/P3 |
改写模板与判别表见 ../writing/introduction-guide-zh.md。
Process Chapter Mainline Checks (--process-chapter)
uv run python -B scripts/analyze_logic.py thesis.tex --process-chapter过程分析章(工业/过程背景第二章"工艺分析 + 全文方法框架"章式)主线专项检查,全部 [Script] 启发式,仅在传入该 flag 时运行(默认行为不变)。默认扫描第 2 章,--section 可覆盖目标章。
章式预判(双信号):目标章的章/节标题须同时命中①过程信号(工艺/流程/过程分析/变量分析) 与②框架信号(总体框架/技术框架/研究方案/总体方案/方案框架)才套用 P-* 检查;否则只输出一条 Info("若为方法+实验章式请走方法章规则"),不强套过程分析章检查(方法章常见的"问题描述/ 总体框架"节名不再单独触发)。
| Check | Rule | Severity |
|---|---|---|
| P-FLOW | 工艺/过程分析节内无 \ref{fig:...} 流程图引用(工艺章无流程图) | Major/P1 |
| P-DERIVE | 难点/问题节缺工艺特性词 → Major;有特性词但无因果连接(导致/使得/难以/造成…)→ Minor | Major/P1 或 Minor/P2 |
| P-FRAME | 框架节无框架图引用,或未覆盖 ≥2 个方法模块名/后续章指向(框架空泛);"第 X 章"显式章号映射缺失仅 Info(推荐加强项,5/5 范文框架节均不写章号、章号映射惯例放绪论组织结构节,不写亦合规) | Major/P1(缺图/空泛);Info/P3(缺章号映射) |
| P-ORDER | 框架节先于难点/问题节出现(违顺序不变式) | Minor/P2 |
写作规范与正反例见 ../writing/process-chapter-guide-zh.md。
Body-Chapter Stitching & Intro Bridging (default)
- P-PAPER(默认全章,无需 flag):可见正文出现"源论文/小论文/N 篇论文"表述即报(Minor/P2), 逐处报告不截断——这是盲审最直接的拼接铁证,建议改"核心问题/研究内容/本章"。
- 缺承上分级:第 3 章起章引言缺承接时,若章内其余部分出现"第 X 章"依赖线索(复用前章 产出)→ 维持 Major;纯并列章 → 降 Info(并列方法章可不承上,5 篇范文核实)。推荐承接 写法(角色复用句)见
../writing/method-chapter-guide-zh.md。 --first-chapter N:单章文件运行时声明文件内首个\chapter的真实章号,使承上启下 检查按真实章序生效(缺省时单章文件被视为第一正文章,承上检查静默)。跨章检查(承上启下、 章间主线、P-PAPER 全文覆盖)建议在装配 document.tex 上运行。
Thesis Writing Mainline
When the user asks how to rewrite 绪论、方法章节、实验讨论、总结与展望, map the section to:
研究背景 -> 技术瓶颈/研究空白 -> 科学问题 -> 本文方法/章节工作 -> 实验证据 -> 贡献闭合 -> 局限与展望Return paragraph roles and evidence status. Do not invent citations, experiments, or contribution claims.
Transition Signals
| Relation | Chinese | English |
|---|---|---|
| Addition | 此外、进一步 | furthermore, moreover |
| Contrast | 然而、但是 | however, nevertheless |
| Causation | 因此、由此可见 | therefore, consequently |
| Sequence | 首先、随后 | first, subsequently |
Full details: see
../writing/logic-coherence.md