pdf-json-parser

Objective

The goal of this project is to design and implement a Python-based program that can read a PDF file and convert its contents into a well-structured JSON format. Instead of extracting plain text, the program aims to preserve the logical hierarchy of the document such as pages, sections, sub-sections, paragraphs, tables, and charts/images. This structured JSON output can be further used in data analytics, search engines, or knowledge management systems. Requirements

Input & Output

• Input: A PDF file (e.g., financial reports, fact sheets, research papers). • Output: A JSON file that organizes extracted content into a clear structure with types like paragraph, table, chart.

Libraries Used

1. pdfplumber → for extracting text and simple tables.
2. PyMuPDF (fitz) → for extracting images and handling layout.
3. streamlit → for creating a user interface.
4. camelot (optional) → for advanced table extraction.

Methodology

The project follows a modular approach, where each step of PDF parsing is handled separately:

1. Text & Paragraph Extraction o Text is extracted using pdfplumber. o Heuristics are applied to split text into paragraphs by checking line breaks, sentence endings, and spacing.
2. Heading & Section Detection o Font size and capitalization are analyzed to identify sections and sub-sections. o Example: Larger font or uppercase lines are treated as section headers.
3. Table Extraction o Tables are identified using pdfplumber’s built-in table detection. o Optionally, camelot is used for more accurate extraction of complex tables.
4. Chart/Image Extraction o Images (likely charts or figures) are extracted using PyMuPDF. o They are saved as separate PNG files and referenced inside the JSON output.
5. JSON Structuring o Each page is represented as an object in JSON. o Inside each page, content is divided into paragraph, table, or chart. o Section headers are linked to paragraphs when possible.

JSON Structure (Example)

{ "pages": [ { "page_number": 1, "content": [ { "type": "paragraph", "section": "Introduction", "sub_section": "Background", "text": "This project demonstrates how PDF content can be structured..." }, { "type": "table", "section": "Financial Data", "description": null, "table_data": [ ["Year", "Revenue", "Profit"], ["2022", "$10M", "$2M"], ["2023", "$12M", "$3M"] ] }, { "type": "chart", "section": "Performance Overview", "description": "Chart showing revenue growth over years", "image_path": "extracted_images/p1_img_0.png", "width": 1200, "height": 600 } ] } ] }

Conclusion

This project successfully demonstrates how a PDF file can be transformed into a machine-readable JSON format while preserving the document hierarchy and content types. The solution can: • Extract clean text paragraphs, • Detect and store tables, • Capture charts/images for reference, • Organize everything in a structured JSON. Such a system is highly valuable for businesses, researchers, and developers who need to convert unstructured documents into structured data for analytics, automation, and search applications.

OUTPUT