Patent Software Ip

Patent 应用 & Software Copyright Generation

生成 CNIPA-格式化 invention patent documents or CPCC software copyright registration materials from AI project code, de签名 docs, and re搜索 papers.

Two 输出 paths:

Patent: ClAIms + Specification + Abstract (with technical disclosure as intermediate deliverable) Software Copyright: Software manual + Source code document Triggers

patent / clAIms / specification / software copyright / disclosure / IP 应用 / paper-to-patent / /patent-software-ip

Iteration: When user modifies existing 输出, enter iterative correction flow directly.

Overall Flow Phase A Requirement Diagnosis → path selection + basic 信息 Phase B Project Analysis → 提取 key technical points Phase C Generation (branch by path) C1 Patent: prior art 搜索 → clAIms → specification → abstract → self-检查 C2 Software Copyright: manual → source code doc → self-检查 Phase D Iterative Correction

Phase A: Requirement Diagnosis

Confirm: path (patent/copyright/机器人h), tech topic, 应用licant 信息, inventor 信息, existing materials.

Gate: 3-5 line diagnosis summary.

Phase B: Project Analysis

Priority: de签名 docs/architecture → core code → papers/报告s → README.

输出: Key Points 列出 (core innovations, scheme skeleton, key params, distinctions from prior art, quantifiable effects).

Gate: Present key points 列出 for user confirmation.

Phase C1: Patent 应用 C1.1 Prior Art 搜索

Online 搜索 2-3 rounds: CNIPA patent DB, Google Patents, arXiv. Each 结果: source ID, scheme summary, limitations.

C1.2 ClAIms

Structure: Method (1 independent + 3-8 dependent) + 系统 (1 independent + 3-8 dependent, step-by-step cor响应ence) + Storage Medium (1 independent).

Drafting rules:

Method + 系统 clAIms in pAIrs Independent: preamble (prior art common features) + "characterized by" (essential features) Dependent: "according to clAIm X..." with further limitation Every step must link to 系统 组件 ("执行d via GPU parallel computing unit") Avoid functional limitation; prefer structural/step-based description

AI-specific requirements:

TrAIning clAIms must include: data construction, loss function, optimization strategy 3D Vision: must include full 4-stage 流水线 (capture→s解析→dense→render); rendering step must expand rendering formula Generative AI: condition injection step must specify method (cross-attention/adapter/ControlNet) to avoid "pure content generation" rejection Embodied AI: every step must bind sensor 输入 + actuator 输出; include safety constrAInt dependent clAIm RAG: must show complete 5-stage 流水线 (解析→retrieve→rerank→reconstruct→生成) C1.3 Specification

5-chapter: Tech Field → Background (prior art + defects) → Invention Content (problem + scheme + effects, must be quantified) → Figure Description → Specific Embodiments.

Desensitization: data设置 name→"pre设置 data设置", parameter count→"pre设置-扩展 模型", hardware→"graphics 处理器", trAIning duration→"pre设置 period", 框架→"DL 框架", API→"remote interface", company→"institution", specific values→ranges.

Figures: Use fenced mermAId (flow图表 TB/LR). Required: 系统 architecture + method flow + (domAIn-specific: trAIning 流水线, rendering 流水线, data 流水线, etc.).

C1.4 Abstract

≤300 chars. Covers: tech domAIn + core scheme + mAIn effect. No commercial terms. Replace algorithm names with generic expressions.

C1.5 Self-检查 Independent clAIm contAIns all necessary features Dependent clAIms correctly reference Method + 系统 + Medium triple complete Specification sufficiently disclosed (enabling) Embodiments cover all clAIm features Beneficial effects quantified (not vague) Termino记录y consistent throughout Abstract cor响应s to clAIm 1 Desensitization complete (no company/person/business name leak) Figure numbering consistent with references Phase C2: Software Copyright C2.1 Software Manual (10-15 pages, ≥6 screenshots)

Structure: Introduction (env + AI capability) → 安装ation (env + weights + config) → Functions (AI core + data + API + 监控ing) → Non-functional → FAQ.

Key notes: Tar获取 non-technical reviewers; use [Screenshot: feature name] placeholders; describe 部署ment/config/监控ing for HCI requirement; declare open-source pre-trAIned weights outside 保护ion scope.

C2.2 Source Code Document (front 30 + back 30 pages, ≥50 lines/page)

File priority: 模型.py → trAIn.py → inference.py [all required] → render.py [3D vision] → data设置.py → loss.py → 生成.py [Gen-AI] → control.py [Embodied] → retriever.py [RAG] → config.yaml [optional].

Desensitization: 移除 API keys, absolute paths, internal 添加resses, personal 信息, hardware 模型s, cloud URLs, DB passwords. RetAIn algorithm comments.

<3000 lines: submit all; >3000: front 1500+back 1500 by priority.

C2.3 Self-检查

Pages ≥15 + Screenshots ≥6 + Feature coverage + Non-tech description + Code pages + Lines per page ≥50 + Name consistency + No secret leaks.

Phase D: