DokuWiki

# 📁 Patent Folder Ingest & Validation Module (MVP Part 1)

## 🎯 Purpose

To scan a local directory of patent folders, validate the file structure and content of each, and generate a structured JSON representation of each patent to be used for search, tagging, and future analysis.

—

## 🧩 What This Module Does

### 1. Folder Scanning

- Recursively scans a selected root directory. - Identifies each patent folder by name format:

`[patent number] [patent name]`  
e.g. `US_10101845_B2 Touch Activated Wireless Poster`

—

### 2. Expected Files Per Folder

Each folder may contain up to six files:

- Missing files are logged. - Optional files improve downstream functionality but aren't required for basic ingestion.

—

### 3. File Content Verification

Without using LLMs (yet), we validate content using heuristics:

- `Bibliographic_Data.md`: Look for fields like Title, Inventor(s), Applicant, Filing Date, etc. - `Claims.md`: Should contain numbered claims (e.g., `1.` or `Claim 1.`). If missing or poorly formatted, flag for review. - `Description.md`: Should contain common patent section headers (e.g., `Background`, `Summary`, `Embodiments`). - `Legal_Events.md`: Check for date+event pairings or relevant keywords (e.g., `Granted`, `Expired`, `Lapsed`, `Renewed`). - `Patent_Family.md`: Should contain one or more patent numbers.

—

### 4. Canonical Output: JSON per Patent

Each valid folder is converted into a JSON object:

```json {

"patent_number": "US_10101845_B2",
"patent_name": "Touch Activated Wireless Poster",
"folder_path": "/patents/US_10101845_B2 Touch Activated Wireless Poster",
"files_present": [
  "Bibliographic_Data.md",
  "Claims.md",
  "Description.md",
  "Patent_Family.md",
  "US_10101845_B2.pdf"
],
"validation_flags": {
  "missing_files": ["Legal_Events.md"],
  "format_warnings": {
    "Claims.md": "Claims unnumbered or improperly formatted"
  }
},
"extracted_metadata": {
  "priority_date": "2014-06-27",
  "filing_date": "2015-06-25",
  "inventors": ["Kate Stone"],
  "assignee": "Novalia Ltd",
  "status": "Granted",
  "family_id": "GB201411234"
}

} ```

This JSON is stored either in memory, written to disk, or passed directly to a backend database or service layer.

—

### 5. Validation Report

A report is generated showing:

- Total patents scanned - Valid patents vs invalid - Missing or malformed files - Common error types (e.g., inconsistent claim formatting)

This is output as a `.json` or `.log` file and optionally shown in the UI.

—

## 🏗️ Future Extensibility (Not part of MVP 1)

- LLM summarization of claims and descriptions - Natural-language tagging and categorization - Full patent family visualizations - File repair / normalization tools - Integration with Espacenet, USPTO, or Google Patents APIs

—

## 💻 Technical Summary

—

## 🧠 Summary

This foundational module ingests a manually curated set of patent folders, verifies that they are complete and well-structured, and produces standardized, reliable JSON files for each. These form the basis for downstream tools: search, summaries, tagging, analytics, and UI presentation.