README.md

metadata

license: apache-2.0
pretty_name: SEC Filings Dataset Parser
tags:
  - sec
  - edgar
  - financial-documents
  - document-parsing
  - markdown

SEC Filings Dataset Parser

This repository contains the core parser used to convert SEC EDGAR filings into layout-faithful Markdown-style text for downstream dataset construction and evaluation.

Contents

• sec_parser/sec_parser.py: main parser implementation
• sec_parser/special_chars.py: special-character normalization tables
• sec_parser/hardcodes.py: filing cleanup hardcodes
• sec_parser/config.py: parser configuration
• pdf_table_fastpath.py, table_ocr_backends.py, mistral_pdf_ocr_overlay.py: PDF/table OCR helper code
• test_sec_parser.py: smoke/regression tests
• Dockerfile, Dockerfile.linux, requirements.txt: reproducible environment files

Quick Start

pip install -r requirements.txt
python -m unittest test_sec_parser.py
python sec_parser/sec_parser.py path/to/filing.txt

Some PDF/OCR paths require external OCR credentials and browser rendering dependencies.