Dorsal PDF Extractor

Fast, deterministic PDF extraction using pdfium. Get native text and bounding boxes without the overhead of heavy AI layout models. Tesseract OCR fallback for scanned pages.

Features

• Native Extraction: Extracts text and precise spatial coordinates (bounding boxes) directly from the PDF document.
• Tesseract OCR Fallback: Built-in optical character recognition (OCR) fallback for scanned or empty pages (requires pytesseract)
• Schema Compliant: Outputs standardized JSON matching the open/document-extraction schema.
• Interactive HTML Export: Outputs can be exported to interactive HTML wireframes via the Dorsal CLI (requires dorsalhub-adapters).

Quick Start

Run the model directly against a local PDF file:

dorsal run dorsalhub/pdf-extractor ./document.pdf

Configuration Options

In the CLI, you can pass options to the model using the --opt (or -o) flag.

Example: Run the extractor with OCR fallback enabled for a French language document:

dorsal run github:dorsalhub/pdf-extractor ./document.pdf --opt use_ocr=true --opt ocr_language=fra

Options:

• password (default: null): Password for decrypted protected PDFs.
• strict (default: false): Toggle strict parsing mode for PDF processing.
• use_ocr (default: false): Enable Tesseract OCR fallback for pages where no native text tokens are detected.
• ocr_language (default: "eng"): The language code to use for the OCR fallback engine.

Output Formats & Exporting

By default, the CLI outputs a validated JSON record to the current working directory.

You can export to other formats right from the CLI. For example, exporting to HTML:

dorsal run github:dorsalhub/pdf-extractor ./document.pdf --export=html

Output

This model produces a file annotation conforming to the Open Validation Schemas Document Extraction schema:

• Schema ID: open/document-extraction (v0.5.0)
• Key Fields:
• extraction_type: Indicates the type of extraction (e.g., boxes, text, or mixed).
• unit: Geometric unit, set to per_mille for standardized relative spatial mapping.
• page_width / page_height: Absolute dimensions of the pages in pixels.
• blocks: An array of elements containing the block_type, text, page_number, and box coordinates.

Development

Running Tests

This repository uses pytest for integration testing.

pip install -e .
pytest

License

This project is licensed under the Apache 2.0 License.