Breno's picture
Create README.md
8b4bba8 verified

Dataset: Legal Documents from STJ for Jurimetrics Research

Dataset Overview

This dataset contains legal documents from the Superior Tribunal de Justiça (STJ), designed for research in jurimetrics, automatic text summarization, and retrieval-augmented generation (RAG). The dataset focuses on the challenges posed by hierarchical structures, legal vocabulary, ambiguity, and citations in legal texts.

Contents

The dataset includes:

  • Ementas (Summaries): Concise summaries of legal decisions.
  • Document Types: Classified by resource types (e.g., appeals, decisions, opinions).
  • Tokens Count: Pre-calculated token counts for analyzing document lengths.
  • Metadata: Additional attributes such as document dates, involved parties, and court sections.

Dataset Features

Feature Name Description Data Type
id Unique identifier for the document. String
type_of_resource Type of the legal document (e.g., appeal). String
ementa Summary of the legal decision. String
full_text Full content of the legal decision. String
token_count Number of tokens in the document summary. Integer
date Date of the decision (YYYY-MM-DD). Date
metadata Additional information (parties, sections). JSON Object

Dataset Usage

This dataset can be used for tasks such as:

  1. Automatic Text Summarization: Evaluating algorithms for generating or refining summaries.
  2. Document Classification: Identifying the type or category of legal documents.
  3. Retrieval-Augmented Generation (RAG): Improving legal text retrieval and contextual generation.
  4. Token Analysis: Studying the distribution and challenges of token lengths in legal summaries.

Data Source

The data is sourced from publicly available legal decisions on the Superior Tribunal de Justiça (STJ). Preprocessing steps were applied to ensure data consistency and usability for machine learning models.

Note: Ensure compliance with ethical and legal considerations regarding the use of public legal documents.

How to Load the Dataset

Using the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("your-username/stj-legal-documents")
print(dataset)