ipo-text / README.md
siddharthlohani's picture
Update README.md
c82882b verified
metadata
license: cc-by-4.0
task_categories:
  - text-classification
  - sentence-similarity
  - document-question-answering
size_categories:
  - 100K<n<1M
language:
  - en
tags:
  - finance
  - nlp
  - text
configs:
  - config_name: default
    data_files:
      - split: train
        path:
          - metadata.jsonl
dataset_info:
  features:
    - name: company
      dtype: string
    - name: company_tickers
      sequence: string
    - name: cik
      dtype: string
    - name: sic
      dtype: string
    - name: office
      dtype: string
    - name: industry
      dtype: string
    - name: exchanges
      sequence: string
    - name: filing_type
      dtype: string
    - name: filing_date
      dtype: string
    - name: filing_accession_number
      dtype: string
    - name: filing_processed
      dtype: bool
    - name: parsed_toc_keys
      dtype: string
    - name: parsed_prospectus_summary_valid
      dtype: bool
    - name: parsed_prospectus_summary
      dtype: string
    - name: parsed_prospectus_summary_word_count
      dtype: int64
    - name: parsed_risk_factors_valid
      dtype: bool
    - name: parsed_risk_factors
      dtype: string
    - name: parsed_risk_factors_word_count
      dtype: int64
    - name: parsed_legal_matters_valid
      dtype: bool
    - name: parsed_legal_matters
      dtype: string
    - name: parsed_legal_matters_word_count
      dtype: int64
    - name: filing_url
      dtype: string
    - name: filing_images
      sequence: string
pretty_name: SEC IPO Filings Datase

SEC IPO Filings Dataset

A large-scale, comprehensive dataset of 100,000+ filings (S-1 and F-1 filings) filed with the SEC EDGAR system, spanning 1994–2026 and over 20,000 unique registrants.

Every filing has been downloaded and then parsed using the IPO-Mine Python Package. We have extracted three common sections found in these documents (Prospectus Summary, Risk Factors, Legal Matters), and then used an LLM classifier to group them into three categories. For this dataset, we have only included the parsed sections labeled as "Safe to Use" as valid.

Software Pipeline

Dataset Construction Pipeline

Citation

The dataset is a product of the following research. If you use this dataset in your work, please cite:

@misc{galarnyk2026ipomine,
  title  = {IPO-Mine: A Toolkit and Dataset for Section-Structured Analysis of Long, Multimodal IPO Documents},
  author = {Galarnyk, Michael and Lohani, Siddharth and Nandi, Sagnik and Patel, Aman and Kannan, Vidhyakshaya and Banerjee, Prasun and Routu, Rutwik and Ye, Liqin and Hiray, Arnav and Somani, Siddhartha and Chava, Sudheer},
  year   = {2026},
  url    = {[https://huggingface.co/datasets/gtfintechlab/ipo-images](https://huggingface.co/datasets/gtfintechlab/ipo-images)},
  note   = {Preprint/Working Paper}
}