# AI 可学习的长篇叙事拆解方法论 **版本**: v1.0 **日期**: 2025-02-17 **目标**: 将优质故事逆向拆解成 AI 可学习的思考步骤 --- ## 核心理念 ### 方法论的三个层次 ``` 宏观层（故事整体）: MICE线程管理 + Save the Cat节拍 ↓ 中观层（章节段落）: 起承转合 + 爽点钩子布局 ↓ 微观层（场景细节）: Scene-Sequel结构 ``` ### 逆向工程的目标 不是简单地"从大纲到正文"，而是提取**作者的思考过程（Chain of Thought）**： 1. **结构决策**: 为什么在这个位置安排这个事件？ 2. **情感设计**: 如何通过情节控制读者情绪？ 3. **节奏把控**: 何时加速，何时放缓？ 4. **爽点规划**: 如何设计和分布爽点？ 5. **钩子布置**: 如何制造和满足期待？ --- ## 一、宏观层拆解：故事架构 ### 1.1 MICE 线程识别与标注 #### 标注维度 ```json { "mice_threads": [ { "thread_id": "M001", "type": "Milieu", "priority": 2, "opening": { "chapter": 1, "position": 0.01, "trigger": "主角进入新世界的标志性事件", "description": "许七安在监狱中醒来，发现穿越到大奉王朝" }, "development": [ { "chapter": 2, "event": "了解京兆府衙门的运作", "world_building": "官场生态、权力结构" }, { "chapter": 7, "event": "进入京城繁华街道", "world_building": "社会风貌、文化特色" } ], "closing": { "chapter": null, "position": null, "trigger": "主角离开或完全融入这个世界", "status": "未关闭" }, "nesting_level": 1 }, { "thread_id": "I001", "type": "Idea", "priority": 3, "opening": { "chapter": 1, "position": 0.15, "question": "现代刑侦方法能否在古代仙侠世界有效？", "description": "许七安思考如何用现代知识破案" }, "development": [ { "chapter": 4, "partial_answer": "可以，通过逻辑推理发现破绽", "evidence": "时间和重量的矛盾" }, { "chapter": 5, "partial_answer": "可以，通过化学知识制造证据", "evidence": "金属钠实验" } ], "closing": { "chapter": 6, "position": 0.06, "answer": "完全可以，现代科学思维在古代同样有效", "status": "阶段性关闭" }, "nesting_level": 3 }, { "thread_id": "C001", "type": "Character", "priority": 2, "opening": { "chapter": 1, "position": 0.05, "dissatisfaction": "不满于即将被流放的命运", "description": "许七安从现代警察变成古代阶下囚，身份认同危机" }, "development": [ { "chapter": 1, "stage": "求生欲驱动", "internal_conflict": "恐惧vs希望" }, { "chapter": 4, "stage": "展现能力", "internal_conflict": "自我怀疑vs自信" }, { "chapter": 7, "stage": "融入家庭", "internal_conflict": "陌生感vs归属感" }, { "chapter": 10, "stage": "确立价值观", "internal_conflict": "摸鱼vs坚持正义" } ], "closing": { "chapter": null, "position": null, "resolution": "接受新身份或改变命运", "status": "未关闭" }, "nesting_level": 2 }, { "thread_id": "E001", "type": "Event", "priority": 1, "opening": { "chapter": 1, "position": 0.03, "disruption": "税银失窃，朝野震动，许家面临灭顶之灾", "description": "世界的正常秩序被打破" }, "development": [ { "chapter": 2, "escalation": "发现'妖物作祟'，案件更加复杂" }, { "chapter": 4, "turning_point": "许七安提出新思路，案件出现转机" }, { "chapter": 5, "climax": "实验成功，真相大白" } ], "closing": { "chapter": 6, "position": 0.06, "resolution": "许家获释，秩序恢复", "new_order": "许七安地位提升，获得认可", "status": "已关闭" }, "nesting_level": 4 } ], "nesting_structure": "M[ I[ C[ E[ ]E ]C ]I ]M", "validation": "valid" } ``` #### 思考过程提取 **作者的MICE决策**: 1. **选择主线程**: 为什么选择Event作为最外层？ - 答：税银案提供了明确的时间压力和高风险 - 思考：Event线程适合快节奏开局，立即抓住读者 2. **嵌套顺序**: 为什么Character在Event内部？ - 答：主角的成长依赖于解决事件 - 思考：内层线程为外层线程服务 3. **快速关闭**: 为什么第6章就关闭Event线程？ - 答：避免拖沓，保持节奏 - 思考：网文需要快速满足，然后开启新线程 --- ### 1.2 Save the Cat 节拍定位 #### 标注维度 ```json { "story_length_words": 1000000, "beats": [ { "beat_number": 1, "beat_name": "Opening Image", "target_position_percent": 1, "actual_position_percent": 0.5, "actual_chapter": 1, "actual_word_count": 5000, "description": "许七安在监狱中醒来，面临流放", "function": "展示主角变化前的状态", "emotional_state": "绝望、恐惧", "world_state": "混乱、危机", "author_intent": "立即建立紧张感，倒计时压力" }, { "beat_number": 2, "beat_name": "Theme Stated", "target_position_percent": 5, "actual_position_percent": 4, "actual_chapter": 1, "actual_word_count": 40000, "description": "许七安思考：'现代刑侦手段能否在这个世界有效？'", "function": "暗示主题：科学vs玄学，现代vs古代", "theme": "理性思维的力量", "author_intent": "埋下核心冲突的种子" }, { "beat_number": 4, "beat_name": "Catalyst", "target_position_percent": 10, "actual_position_percent": 8, "actual_chapter": 1, "actual_word_count": 80000, "description": "堂弟许新年带来卷宗", "function": "提供破局的可能性", "emotional_shift": "从绝望到希望萌生", "author_intent": "给主角（和读者）一线生机" }, { "beat_number": 6, "beat_name": "Break into Two", "target_position_percent": 20, "actual_position_percent": 15, "actual_chapter": 2, "actual_word_count": 150000, "description": "许七安决定用现代推理破案", "function": "主角主动选择，进入新阶段", "commitment": "从被动等死到主动求生", "author_intent": "建立主角的主动性" }, { "beat_number": 8, "beat_name": "Fun and Games", "target_position_percent": 35, "actual_position_percent": 30, "actual_chapter": 4, "actual_word_count": 300000, "description": "展示推理过程，化学实验", "function": "前提的承诺：现代知识碾压古代", "reader_expectation": "看主角如何用现代知识装逼", "author_intent": "密集爽点，满足读者期待" } ], "emotional_arc": [ {"position_percent": 0, "level": 2, "state": "绝望"}, {"position_percent": 10, "level": 4, "state": "希望"}, {"position_percent": 20, "level": 5, "state": "决心"}, {"position_percent": 35, "level": 8, "state": "兴奋"}, {"position_percent": 50, "level": 9, "state": "高潮"}, {"position_percent": 60, "level": 6, "state": "平静"}, {"position_percent": 75, "level": 3, "state": "低谷"}, {"position_percent": 85, "level": 7, "state": "顿悟"}, {"position_percent": 100, "level": 10, "state": "满足"} ] } ``` #### 思考过程提取 **作者的节拍决策**: 1. **提前触发**: 为什么Catalyst在8%而非10%？ - 答：网文需要更快的节奏 - 思考：根据媒介调整标准节拍位置 2. **压缩Debate**: 为什么几乎没有犹豫阶段？ - 答：时间压力（两天后流放）不允许犹豫 - 思考：用外部压力替代内心犹豫 3. **提前高潮**: 为什么第5-6章就达到高潮？ - 答：快速满足读者，然后开启新故事 - 思考：百万字长篇需要多个高潮循环 --- ## 二、中观层拆解：章节设计 ### 2.1 起承转合结构 #### 标注维度 ```json { "chapter_id": "chapter_004", "chapter_title": "是时候表演真正的技术了", "word_count": 3500, "structure": { "起": { "position": "0-500字", "function": "建立场景和冲突", "content": "许七安求见府尹，被狱卒阻拦", "hook": "他有重要线索能破案", "emotional_setup": "紧张、期待" }, "承": { "position": "500-2000字", "function": "展开冲突，增加难度", "content": "陈府尹暴躁，几乎要杖责；许七安展示推理", "escalation": [ "府尹不信", "要打板子", "中年男人制止", "开始推理" ], "emotional_development": "从紧张到自信" }, "转": { "position": "2000-3000字", "function": "反转，超出期待", "content": "用数学计算和时间推理揭示破绽", "reversal": "官员们从怀疑到震惊", "shuang_point": "智商碾压，古代人算不出来", "emotional_peak": "爽、兴奋" }, "合": { "position": "3000-3500字", "function": "解决当前冲突，引出新钩子", "content": "官员们认可推理，但需要证据", "resolution": "获得做实验的机会", "new_hook": "能否成功制造金属钠？", "emotional_closure": "满足但期待下一步" } }, "author_thinking": { "why_this_structure": "快速推进，不拖泥带水", "pacing_decision": "承部分占比最大，充分展示推理过程", "hook_strategy": "每个'合'都要引出新'起'" } } ``` --- ### 2.2 爽点设计与分布 #### 爽点类型库 ```json { "shuang_point_types": [ { "type_id": "SP001", "type_name": "智商碾压", "mechanism": "主角用知识/智慧碾压对手", "setup_requirements": [ "建立对手的无知或错误认知", "主角掌握对手不知道的信息", "有旁观者见证" ], "payoff_elements": [ "展示知识的过程", "对手的震惊反应", "旁观者的惊叹" ], "intensity_factors": { "low": "小知识点，对手轻微惊讶", "medium": "重要知识，对手震惊", "high": "颠覆性知识，对手三观崩塌" }, "example": { "chapter": 4, "setup": "官员们认为税银被妖物劫走", "payoff": "许七安用数学计算证明时间和重量矛盾", "reaction": "'竟然是这样！'中年男人猛的站起身", "intensity": "high" } }, { "type_id": "SP002", "type_name": "打脸", "mechanism": "对手嘲讽→主角反击→对手后悔", "setup_requirements": [ "对手的轻视或嘲讽", "主角隐藏实力或被误解", "有明确的对抗场景" ], "payoff_elements": [ "主角展示真实实力", "对手的震惊和尴尬", "身份或地位的反转" ], "intensity_factors": { "low": "小配角被打脸", "medium": "重要配角被打脸", "high": "大反派或权威被打脸" }, "example": { "chapter": 4, "setup": "陈府尹怒喝'胡说八道，拖下去杖责'", "payoff": "许七安的推理让府尹无话可说", "reaction": "府尹从暴躁到沉默", "intensity": "medium" } }, { "type_id": "SP003", "type_name": "升级", "mechanism": "能力提升，实力飞跃", "setup_requirements": [ "建立当前实力的局限", "铺垫升级的可能性", "制造升级的契机" ], "payoff_elements": [ "升级的仪式感", "新能力的展示", "他人的认可" ], "intensity_factors": { "low": "小幅提升", "medium": "境界突破", "high": "质的飞跃" } }, { "type_id": "SP004", "type_name": "装逼", "mechanism": "展示实力，引发惊叹", "setup_requirements": [ "隐藏实力", "有展示的场景", "有旁观者" ], "payoff_elements": [ "实力展示的过程", "旁观者的震惊", "主角的淡定" ] }, { "type_id": "SP005", "type_name": "获得", "mechanism": "获得宝物/技能/资源", "setup_requirements": [ "建立获得物的价值", "制造获得的难度", "展示获得的过程" ], "payoff_elements": [ "获得的惊喜", "获得物的描述", "他人的羡慕" ] }, { "type_id": "SP006", "type_name": "碾压", "mechanism": "实力差距，轻松获胜", "setup_requirements": [ "建立敌人的强大", "主角实力更强", "有战斗场景" ], "payoff_elements": [ "轻松获胜的过程", "敌人的绝望", "旁观者的震惊" ] } ] } ``` #### 爽点分布规划 ```json { "chapter_id": "chapter_004", "word_count": 3500, "shuang_points": [ { "sp_id": "SP_C4_001", "type": "打脸", "position_word": 500, "position_percent": 14, "intensity": "medium", "setup": { "start_word": 300, "content": "陈府尹怒喝'胡说八道'", "setup_length": 200 }, "payoff": { "start_word": 500, "content": "中年男人制止，许七安开始推理", "payoff_length": 300 }, "reaction": { "character": "陈府尹", "reaction": "从暴躁到沉默" } }, { "sp_id": "SP_C4_002", "type": "智商碾压", "position_word": 2000, "position_percent": 57, "intensity": "high", "setup": { "start_word": 1500, "content": "官员们算不出白银重量", "setup_length": 500 }, "payoff": { "start_word": 2000, "content": "许七安秒算：九千三百七十五斤", "payoff_length": 200 }, "reaction": { "character": "中年男人", "reaction": "猛的站起身，'竟然是这样！'" }, "author_comment": "速算能力有点Low啊，你们这群古代人" }, { "sp_id": "SP_C4_003", "type": "智商碾压", "position_word": 2800, "position_percent": 80, "intensity": "high", "setup": { "start_word": 2200, "content": "官员们没注意到时间矛盾", "setup_length": 600 }, "payoff": { "start_word": 2800, "content": "许七安指出时间推理的破绽", "payoff_length": 300 }, "reaction": { "character": "全场", "reaction": "恍然大悟" } } ], "shuang_point_density": 0.86, "shuang_point_distribution": "前期铺垫，中后期密集爆发", "author_thinking": { "why_3_points": "一章3个爽点，保持密度但不过度", "why_this_order": "先打脸建立主角地位，再智商碾压展示能力", "intensity_curve": "medium → high → high，逐步升级" } } ``` --- ### 2.3 钩子布置策略 #### 钩子类型库 ```json { "hook_types": [ { "type_id": "H001", "type_name": "悬念钩子", "mechanism": "提出问题，延迟回答", "placement": "章节末尾", "resolution_timing": "1-3章内", "example": { "chapter": 4, "hook": "能否成功制造金属钠？", "resolution_chapter": 5 } }, { "type_id": "H002", "type_name": "冲突钩子", "mechanism": "制造冲突，不立即解决", "placement": "章节中部或末尾", "resolution_timing": "下一章或几章后", "example": { "chapter": 10, "hook": "王捕头想找替死鬼，许七安能否坚持原则？", "resolution_chapter": 11 } }, { "type_id": "H003", "type_name": "秘密钩子", "mechanism": "暗示秘密，逐步揭示", "placement": "任意位置", "resolution_timing": "长期铺垫", "example": { "chapter": 1, "hook": "这个世界有儒道佛妖术士，主角如何在其中生存？", "resolution_chapter": "持续展开" } }, { "type_id": "H004", "type_name": "承诺钩子", "mechanism": "承诺未来的爽点", "placement": "任意位置", "resolution_timing": "中长期", "example": { "chapter": 6, "hook": "许七安获得认可，未来会有什么发展？", "resolution_chapter": "后续章节" } }, { "type_id": "H005", "type_name": "危机钩子", "mechanism": "制造紧迫感", "placement": "开篇或关键节点", "resolution_timing": "短期", "example": { "chapter": 1, "hook": "两天后流放，如何自救？", "resolution_chapter": 6 } } ] } ``` #### 钩子布置规划 ```json { "chapter_id": "chapter_004", "hooks": [ { "hook_id": "H_C4_001", "type": "悬念钩子", "position": "章节开头", "position_word": 50, "content": "许七安有重要线索，能否说服官员？", "setup_elements": [ "许七安求见", "狱卒阻拦", "坚持要见" ], "tension_level": "medium", "resolution": { "chapter": 4, "position_word": 500, "outcome": "成功见到官员" } }, { "hook_id": "H_C4_002", "type": "冲突钩子", "position": "章节中部", "position_word": 300, "content": "陈府尹要杖责许七安，能否化解？", "setup_elements": [ "府尹暴躁", "怒喝'胡说八道'", "命令拖下去" ], "tension_level": "high", "resolution": { "chapter": 4, "position_word": 500, "outcome": "中年男人制止" } }, { "hook_id": "H_C4_003", "type": "悬念钩子", "position": "章节末尾", "position_word": 3400, "content": "推理虽然有道理，但需要证据，能否成功实验？", "setup_elements": [ "推理完成", "官员认可", "但需要证据" ], "tension_level": "high", "resolution": { "chapter": 5, "position_word": 2000, "outcome": "实验成功" } } ], "hook_density": 0.86, "hook_strategy": "章首钩子吸引，章中钩子维持，章末钩子延续", "author_thinking": { "why_3_hooks": "保持持续吸引力", "resolution_timing": "前两个快速满足，最后一个延续到下章", "tension_management": "逐步升级紧张感" } } ``` --- ## 三、微观层拆解：场景结构 ### 3.1 Scene-Sequel 详细标注 ```json { "scene_id": "S_C4_001", "chapter": 4, "position_word": 0, "length_word": 800, "type": "Scene", "structure": { "goal": { "character": "许七安", "want": "见到府尹，提供破案线索", "why": "这是唯一的自救机会", "stakes": "失败则流放边陲", "clarity": "非常明确", "reader_awareness": "读者完全知道目标" }, "conflict": { "obstacles": [ { "type": "人物阻碍", "source": "狱卒", "description": "不相信许七安有线索，阻拦", "intensity": "low" }, { "type": "时间压力", "source": "案件紧急", "description": "官员们正在商议，可能随时离开", "intensity": "medium" } ], "escalation": "从狱卒阻拦到差点被打", "turning_point": "许七安坚持，狱卒妥协" }, "disaster": { "outcome": "成功见到官员，但陈府尹暴躁，要杖责", "type": "部分成功+新危机", "worse_than_expected": true, "description": "虽然见到了，但差点被打", "emotional_impact": "紧张升级", "leads_to_sequel": true } }, "author_thinking": { "goal_design": "明确且紧迫，读者立即理解", "conflict_design": "层层递进，不是一次性解决", "disaster_design": "部分成功制造新冲突，保持节奏" } } ``` ```json { "sequel_id": "SQ_C4_001", "chapter": 4, "position_word": 800, "length_word": 400, "type": "Sequel", "structure": { "reaction": { "character": "许七安", "emotional_response": "紧张但保持冷静", "physical_response": "深呼吸，整理思路", "duration": "短暂", "description": "面对暴躁的府尹，许七安强迫自己冷静" }, "dilemma": { "options": [ { "option": "直接说出推理", "pros": "快速切入主题", "cons": "可能被认为狂妄，继续激怒府尹" }, { "option": "先道歉，再慢慢说", "pros": "缓和气氛", "cons": "可能失去说话机会" }, { "option": "请求中年男人支持", "pros": "借助权威", "cons": "显得依赖他人" } ], "internal_conflict": "如何在不激怒府尹的情况下展示推理", "time_pressure": "府尹随时可能下令", "stakes": "这是最后的机会" }, "decision": { "choice": "直接展示推理，用事实说话", "reasoning": "时间紧迫，没有迂回的余地", "confidence_level": "高", "becomes_next_goal": "说服官员相信推理", "description": "许七安决定直接提出时间和重量的矛盾" } }, "author_thinking": { "reaction_design": "短暂但必要，展示主角的冷静", "dilemma_design": "真实的选择困境，增加代入感", "decision_design": "果断决策，符合主角性格，推动情节" } } ``` --- ### 3.2 对话设计 ```json { "dialogue_id": "D_C4_001", "scene_id": "S_C4_001", "participants": ["许七安", "陈府尹", "中年男人"], "function": "展示冲突和推理", "structure": { "opening": { "speaker": "许七安", "line": "大人，卑职有重要线索", "function": "提出目标", "subtext": "我能帮你们破案" }, "conflict": { "speaker": "陈府尹", "line": "胡说八道，来人，拖下去，杖责两百", "function": "制造障碍", "subtext": "我不相信你" }, "turning_point": { "speaker": "中年男人", "line": "等等，你说说看", "function": "化解危机，给机会", "subtext": "我愿意听" }, "展示": { "speaker": "许七安", "lines": [ "十五万两白银，重几斤？", "九千三百七十五斤", "卯时二刻进城，不可能在辰时一刻抵达广南街" ], "function": "展示推理，智商碾压", "subtext": "我比你们聪明" }, "reaction": { "speaker": "中年男人", "line": "竟然是这样！", "function": "确认爽点效果", "subtext": "你说得对" } }, "dialogue_techniques": { "conflict_escalation": "从质疑到暴怒到制止到认可", "information_reveal": "通过对话逐步揭示推理", "character_voice": "府尹暴躁，中年人冷静，许七安自信", "subtext": "表面是讨论案件，实际是权力和智慧的较量" }, "author_thinking": { "why_this_structure": "快速建立冲突，快速解决，保持节奏", "character_differentiation": "通过对话展现不同性格", "reader_engagement": "让读者跟随推理过程" } } ``` --- ## 四、思考过程（Chain of Thought）提取 ### 4.1 结构决策的思考链 ```json { "decision_type": "结构决策", "question": "为什么第6章就结束税银案？", "thinking_chain": [ { "step": 1, "thought": "税银案是开局的主要冲突", "consideration": "需要快速建立紧张感" }, { "step": 2, "thought": "但这不是整个故事的主线", "consideration": "百万字长篇需要更大的格局" }, { "step": 3, "thought": "如果拖太久，读者会失去耐心", "consideration": "网文读者需要快速满足" }, { "step": 4, "thought": "6章约6万字，已经足够展示主角能力", "consideration": "已经完成了角色建立的任务" }, { "step": 5, "thought": "快速结束可以立即开启新故事", "consideration": "保持新鲜感和节奏" }, { "step": 6, "decision": "第6章结束税银案，第10章开启县衙命案", "reasoning": "快速满足→短暂过渡→新冲突，保持节奏" } ], "alternative_considered": [ { "option": "拖到20章", "rejected_reason": "太慢，读者会流失" }, { "option": "3章结束", "rejected_reason": "太快，无法充分展示主角能力" } ], "final_decision": "6章结束，平衡了展示和节奏" } ``` --- ### 4.2 爽点设计的思考链 ```json { "decision_type": "爽点设计", "question": "如何设计第4章的智商碾压爽点？", "thinking_chain": [ { "step": 1, "thought": "主角的优势是现代知识", "consideration": "要展示现代vs古代的差距" }, { "step": 2, "thought": "数学计算是最直观的差距", "consideration": "古代人算不出来，主角秒算" }, { "step": 3, "thought": "需要建立古代人的无能", "consideration": "先让他们算不出来" }, { "step": 4, "thought": "主角要表现得轻松", "consideration": "增强碾压感" }, { "step": 5, "thought": "需要旁观者的震惊反应", "consideration": "通过他人反应放大爽点" }, { "step": 6, "thought": "可以加入主角的内心吐槽", "consideration": "'速算能力有点Low啊，你们这群古代人'" }, { "step": 7, "decision": "设计'十五万两重几斤'的问题", "reasoning": "简单但古代人算不出，主角秒答，形成强烈对比" } ], "setup_elements": [ "官员们讨论税银重量", "没人能算出来", "许七安被问到" ], "payoff_elements": [ "许七安秒答：九千三百七十五斤", "中年男人猛的站起身", "内心吐槽：你们这群古代人" ], "intensity_factors": { "knowledge_gap": "现代数学vs古代算术", "speed_contrast": "秒算vs算不出", "reaction_amplification": "猛的站起身" } } ``` --- ### 4.3 节奏控制的思考链 ```json { "decision_type": "节奏控制", "question": "如何控制前10章的节奏？", "thinking_chain": [ { "step": 1, "thought": "开局必须快，立即抓住读者", "consideration": "第1章就要有危机和钩子" }, { "step": 2, "thought": "但不能一直快，会让读者疲劳", "consideration": "需要张弛有度" }, { "step": 3, "thought": "第1-5章是第一波高潮（破案）", "consideration": "快节奏，密集爽点" }, { "step": 4, "thought": "第6-9章是过渡期（回家）", "consideration": "慢节奏，日常互动，角色塑造" }, { "step": 5, "thought": "第10章开启新故事（县衙命案）", "consideration": "重新加速，制造新期待" }, { "step": 6, "decision": "快-慢-快的节奏循环", "reasoning": "符合情感曲线，避免单调" } ], "pacing_techniques": { "fast_pacing": [ "短句短段", "密集对话", "快速场景切换", "高频爽点" ], "slow_pacing": [ "长句长段", "细节描写", "日常互动", "角色内心" ], "transition": [ "用钩子连接快慢", "慢节奏中埋伏笔", "快节奏中有喘息" ] }, "rhythm_pattern": "快快快快快-慢慢慢慢-快" } ``` --- ## 五、训练数据格式设计 ### 5.1 多层次标注格式 ```json { "story_id": "dafeng_001", "story_title": "大奉打更人", "total_words": 5000000, "analysis_scope": { "start_chapter": 1, "end_chapter": 10, "word_count": 100000 }, "macro_structure": { "mice_threads": [...], "save_the_cat_beats": [...], "emotional_arc": [...] }, "meso_structure": { "chapters": [ { "chapter_id": "chapter_004", "qichengzhuanhe": {...}, "shuang_points": [...], "hooks": [...] } ] }, "micro_structure": { "scenes": [ { "scene_id": "S_C4_001", "scene_sequel": {...}, "dialogue": {...} } ] }, "author_thinking": { "structure_decisions": [...], "shuang_point_designs": [...], "pacing_controls": [...] } } ``` --- ### 5.2 思考过程（CoT）训练格式 ```json { "task": "设计第4章的智商碾压爽点", "context": { "current_chapter": 4, "previous_events": "许七安获得卷宗，发现疑点", "character_state": "许七安需要说服官员", "story_goal": "破案自救" }, "thinking_process": [ { "step": 1, "question": "主角的优势是什么？", "answer": "现代知识，特别是数学和逻辑", "reasoning": "穿越者的核心优势" }, { "step": 2, "question": "如何展示这个优势？", "answer": "让古代人做不到的事，主角轻松做到", "reasoning": "对比产生爽感" }, { "step": 3, "question": "具体用什么知识点？", "answer": "数学计算：十五万两白银的重量", "reasoning": "简单但古代人算不出，主角秒答" }, { "step": 4, "question": "如何增强爽感？", "answer": "加入旁观者震惊反应和主角内心吐槽", "reasoning": "多角度放大爽点效果" }, { "step": 5, "question": "如何设置铺垫？", "answer": "先让官员们讨论但算不出来", "reasoning": "建立对比，突出主角能力" } ], "final_design": { "setup": "官员们讨论税银重量，无人能算", "payoff": "许七安秒答九千三百七十五斤", "reaction": "中年男人猛的站起身", "internal_comment": "速算能力有点Low啊，你们这群古代人" }, "expected_effect": { "reader_emotion": "爽、兴奋", "character_perception": "主角很聪明", "plot_advancement": "获得官员信任" } } ``` --- ### 5.3 对比学习格式 ```json { "comparison_type": "爽点设计对比", "good_example": { "source": "大奉打更人第4章", "shuang_point": "智商碾压", "setup": "官员们算不出税银重量", "payoff": "许七安秒答", "reaction": "中年男人猛的站起身", "why_good": [ "铺垫充分：先建立古代人的无能", "对比强烈：算不出vs秒答", "反应到位：震惊的肢体语言", "有内心吐槽：增加趣味性" ], "intensity": "high" }, "bad_example": { "setup": "主角说出答案", "payoff": "官员们点头", "reaction": "嗯，有道理", "why_bad": [ "没有铺垫：没有建立对比", "反应平淡：点头太弱", "缺少细节：没有震惊的描写", "没有放大：缺少内心吐槽或旁观者" ], "intensity": "low" }, "key_differences": [ "铺垫的重要性：建立对比", "反应的强度：肢体语言vs口头认可", "细节的丰富度：多角度描写", "趣味性：内心吐槽增加可读性" ] } ``` --- ## 六、实施步骤 ### 6.1 数据生产流程 ``` 步骤1: 选择优质样本 ↓ 步骤2: 宏观层标注（MICE + Save the Cat） ↓ 步骤3: 中观层标注（起承转合 + 爽点钩子） ↓ 步骤4: 微观层标注（Scene-Sequel + 对话） ↓ 步骤5: 提取思考过程（CoT） ↓ 步骤6: 生成对比样本（好vs坏） ↓ 步骤7: 质量验证和迭代 ``` --- ### 6.2 优先级排序 #### 第一优先级（核心环节） 1. **Scene-Sequel 结构识别** - 原因：最基础的叙事单元 - 数据量：每个场景都要标注 - 训练目标：让AI学会场景的因果逻辑 2. **爽点设计思考链** - 原因：网文的核心竞争力 - 数据量：每个爽点都要提取思考过程 - 训练目标：让AI学会设计爽点 #### 第二优先级（结构支撑） 3. **MICE 线程管理** - 原因：宏观结构的骨架 - 数据量：每个故事的主要线程 - 训练目标：让AI学会线程的开启、嵌套、关闭 4. **钩子布置策略** - 原因：保持读者粘性的关键 - 数据量：每章的钩子 - 训练目标：让AI学会制造和满足期待 #### 第三优先级（精细打磨） 5. **对话设计** - 原因：展现角色和推进情节 - 数据量：关键对话场景 - 训练目标：让AI学会写有冲突和信息量的对话 6. **节奏控制** - 原因：整体阅读体验 - 数据量：每章的节奏标注 - 训练目标：让AI学会快慢节奏的切换 --- ### 6.3 本周训练计划 #### Day 1-2: 数据准备 - 选择3-5个优质样本（网文+剧本） - 完成宏观层标注（MICE + Save the Cat） - 生成100个Scene-Sequel标注样本 #### Day 3-4: 核心训练 - 训练Scene-Sequel识别模型 - 训练爽点设计思考链模型 - 生成对比样本（好vs坏） #### Day 5-6: 验证和迭代 - 用模型生成新场景 - 人工评估质量 - 根据反馈调整训练数据 #### Day 7: 总结和规划 - 整理训练结果 - 评估模型效果 - 规划下一步优化方向 --- ## 七、评估标准 ### 7.1 结构完整性 - [ ] MICE线程是否正确嵌套？ - [ ] Save the Cat节拍是否齐全？ - [ ] Scene-Sequel是否形成因果链？ ### 7.2 爽点有效性 - [ ] 爽点类型是否明确？ - [ ] 铺垫是否充分？ - [ ] 反应是否到位？ - [ ] 密度是否合理？ ### 7.3 钩子有效性 - [ ] 钩子是否制造期待？ - [ ] 满足时机是否合理？ - [ ] 是否形成钩子链？ ### 7.4 节奏合理性 - [ ] 快慢是否交替？ - [ ] 情感曲线是否流畅？ - [ ] 是否避免单调？ ### 7.5 思考链质量 - [ ] 思考步骤是否清晰？ - [ ] 推理是否合理？ - [ ] 是否考虑了替代方案？ - [ ] 最终决策是否有充分理由？ --- ## 八、总结 ### 核心创新点 1. **多层次融合**: 结合西方结构理论（Scene-Sequel、MICE、Save the Cat）和中国网文理论（爽点、钩子、起承转合） 2. **思考过程提取**: 不仅标注"是什么"，更重要的是提取"为什么这样设计" 3. **对比学习**: 通过好坏样本对比，让AI理解什么是有效的设计 4. **可验证性**: 每个层次都有明确的验证标准 ### 适用场景 - **百万字长篇**: 通过MICE和Save the Cat管理宏观结构 - **网文创作**: 通过爽点和钩子保持读者粘性 - **交互式生成**: 通过Scene-Sequel确保每个场景的逻辑性 ### 下一步优化方向 1. **自动化标注**: 开发工具辅助标注过程 2. **质量评估**: 建立自动化的质量评估系统 3. **迭代优化**: 根据生成效果持续优化训练数据 4. **领域扩展**: 扩展到更多类型（科幻、言情、历史等） --- **方法论状态**: v1.0 - 可执行 **下一步**: 生成示例拆解数据