文本和概念分析专家 Prompt
来源: GitHub Gist 作者: kevinz 原文链接: https://gist.github.com/kevinz/a191dfd758971bf31207484c34c86f30
English
This is a sophisticated prompt for text analysis and interactive HTML generation. The prompt defines an AI role as a "Text Analysis Expert" with the following core capabilities:
Core Task
When a user sends text, the AI must: 1. Deeply analyze the text content, extracting key concepts, terms, and their interrelationships. 2. Based on the analysis, generate a complete, structurally valid HTML document that helps users understand the original text through interactive and visual means.
The HTML Gold Standard
The prompt itself is implemented as the HTML gold standard — the output must strictly follow the same structure, interaction logic, CSS style, and JavaScript libraries demonstrated in the reference page.
Mandatory Output Format
- Every response must be a complete HTML document from
<!DOCTYPE html>to</html> - The entire HTML document must be wrapped in a Markdown code block (
html ...`
HTML Generation Guidelines
1. Responsive Design: Layout and content must adapt to different screen sizes 2. Mathematics: MathJax (CHTML) must be used for rendering math formulas 3. Code Blocks: Must use <pre><code class="language-LLL"> tags with syntax highlighting and horizontal scrolling 4. Creativity and Flexibility: Leverage HTML, CSS, and JavaScript capabilities to create informative, easy-to-understand explanatory pages
Technical Stack
- Viz.js: For rendering concept relationship graphs (Graphviz-based)
- Panzoom: For interactive zoom/pan on the knowledge graph
- MathJax: For mathematical formula rendering
- Custom CSS: With CSS variables for theming (--rad, --text-pri, --bg-main, --brand, etc.)
Interaction Design
- Terms in the main content area are highlighted and clickable
- Clicking a term shows detailed explanation in a side panel
- A concept relationship graph is rendered below the content
- Graph nodes are interactive — clicking shows related explanations
- Support for full-screen graph viewing with zoom/pan
- Graph layout toggle (LR/TB) and PNG download
- Google search popup for further exploration of concepts
Concept Taxonomy
The prompt organizes concepts into types:
- concept (yellow): Core conceptual definitions
- branch (green): Task branches and sub-processes
- tech (blue): Technical implementation details
- app (red): Application-level features
中文
这是一个用于文本分析和交互式 HTML 生成的复杂提示词。该提示词定义了 AI 角色为"文本分析专家"，具有以下核心能力：
核心任务
当用户发送文本时，AI 必须： 1. 深入分析文本内容，提取关键概念、术语及其相互关系。 2. 基于分析结果，生成一个结构完整、语法有效的 HTML 文档，通过交互式和可视化的方式帮助用户理解原始文本。
HTML 黄金标准
该提示词本身即作为 HTML 黄金标准的实现——输出必须严格遵循参考页面所展示的相同结构、交互逻辑、CSS 风格和 JavaScript 库。
强制输出格式
- 每个回答必须是从
<!DOCTYPE html>到</html>的完整 HTML 文档 - 整个 HTML 文档必须用 Markdown 代码块包裹（
html ...`
技术栈
- Viz.js：渲染概念关系图（基于 Graphviz）
- Panzoom：知识图谱的交互式缩放/平移
- MathJax：数学公式渲染
- 自定义 CSS：使用 CSS 变量实现主题化
交互设计
- 主内容区的术语可高亮点击
- 点击术语在侧边面板展示详细解释
- 内容下方渲染概念关系图
- 图谱节点可交互，点击展示相关解释
- 支持全屏查看图谱，可缩放和平移
- 图谱布局切换（左右/上下）和 PNG 下载
- Google 搜索弹出窗口用于进一步探索概念
概念分类
提示词将概念组织为以下类型：
- concept（黄色）：核心概念定义
- branch（绿色）：任务分支和子流程
- tech（蓝色）：技术实现细节
- app（红色）：应用级功能