Skip to content

Dataset · eddmpython

US and Korean corporate financial reporting data

eddmpython/dartlab-data

A structured dataset of DART and SEC EDGAR reports in parquet format for market analysis.

What's inside

A large-scale corpus ranging from 1 to 10 million records in Korean and English, provided in parquet format. It includes financial statements, annual report texts, stock prices, and macroeconomic indicators for all public Korean companies (approx. 2,700) and major US companies (approx. 1,000) sourced from DART and SEC EDGAR. Data is preserved in its original form without rounding or interpolation and unified for ticker-based lookup.

What you can do with it

  • Training models for table-based question answering (Table QA)
  • Building RAG systems for fundamental analysis of US and Korean stocks
  • Training models for structured data extraction from financial reports
Tasks
Table question answering, Text classification
Languages
Korean, English
Size
1M–10M rows
Downloads this month
195.8K
Likes
24

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("eddmpython/dartlab-data")