Skip to content

gtfintechlab/IPO-Mine

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IPO-Mine: A Toolkit and Dataset for Section-Structured Analysis of Long, Multimodal IPO Documents

YouTube Paper Dataset PyPI - ipo-mine

Dataset Construction Pipelines

Image Dataset Pipeline Text Dataset Pipeline
Description 1 Description 2

Quickstart

Install from PyPI

pip install ipo-mine

Using ipo-mine to Download an IPO Filing (Python API)

from download import IPODownloader, Company

downloader = IPODownloader(
    email="example@gmail.com",
    company="Your Example Organization"
)

company = Company.from_ticker("SNOW")

company_filings = downloader.download_ipo(
    company,
    limit=1,
    save_filing=True,
    save_images=False,
    verbose=True
)

filing = company_filings.filings[0]

Parsing the Table of Contents

results = parser.parse_company(
    ticker="SNOW",
    validate=False
)

CLI Usage

You can use the command-line interface to download and parse filings without writing Python code.

Download

Download the latest S-1 filing for a company:

ipo-mine download SNOW --email your@email.com --org "Your Org"

Options:

  • --limit N: Download previous N filings (default: 1)
  • --images: Download and extract images from the filing
  • --all: Download all available IPO filings for the ticker

Parse

Parse a downloaded filing into section-specific files:

ipo-mine parse SNOW

Options:

  • --validate: Enable LLM-based validation of extracted sections
  • --provider: LLM provider (anthropic, openai, google, huggingface)
  • --mode: Validation mode (binary, likert)

Validate

Run LLM validation on existing parsed text files to check for truncation or completeness.

ipo-mine validate SNOW --provider anthropic

Supported Providers

You can choose from the following providers (requires API keys):

Provider Argument Env Variable
Anthropic (Claude) --provider anthropic ANTHROPIC_API_KEY
OpenAI (GPT-4o) --provider openai OPENAI_API_KEY
Google (Gemini) --provider google GOOGLE_API_KEY
HuggingFace --provider huggingface HUGGINGFACE_API_KEY

Validation Modes

  • Binary (--mode binary): Returns "Yes" (Valid) or "No" (Truncated/Incomplete). Default.
  • Likert (--mode likert): Returns a confidence score from 1 (Incomplete) to 5 (Complete).

Authentication

The CLI will look for API keys in this order:

  1. Command Line Argument: --api-key "sk-..."
  2. Environment Variable: e.g., export OPENAI_API_KEY="sk-..."
  3. Interactive Prompt: If neither is found, the CLI will securely prompt you to enter the key (input is hidden).

Examples

Validate using OpenAI with Likert scale:

ipo-mine validate TSLA --provider openai --mode likert

Validate using Google Gemini with explicit key:

ipo-mine validate TSLA --provider google --api-key "your-api-key"

Notes

  • The SEC requires a descriptive User-Agent. Provide a real organization name and your email.
  • download_ipo returns a CompanyFilings object; use company_filings.filings[0] to pass a Filing into the parser.
  • The parser automatically chooses HTML or text parsing based on the filing URL.

Citation

If you use IPO-Mine, please cite:

@misc{galarnyk2026ipominetoolkitdatasetsectionstructured,
      title={IPO-Mine: A Toolkit and Dataset for Section-Structured Analysis of Long, Multimodal IPO Documents}, 
      author={Michael Galarnyk and Siddharth Lohani and Vidhyakshaya Kannan and Sagnik Nandi and Aman Patel and Liqin Ye and Arnav Hiray and Rutwik Routu and Prasun Banerjee and Siddhartha Somani and Sudheer Chava},
      year={2026},
      eprint={2605.28714},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2605.28714}, 
}

About

A toolkit and dataset for multimodal IPO filing analysis.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages