Brazilian Portuguese Legal Documents with Relevance Judgments

JurisTCU contains 16,045 legal documents from the Brazilian Federal Court of Accounts. It is a dataset for Legal Information Retrieval in Brazilian Portuguese, providing query relevance judgments to support the evaluation of legal search systems.

Use Cases

• Train and evaluate information retrieval models using the provided query relevance judgments (qrels) on Brazilian Portuguese legal text.
• Benchmark the performance of multilingual or Portuguese language models on the task of legal document retrieval.
• Analyze the characteristics of jurisprudence from the Brazilian Federal Court of Accounts (TCU) for legal NLP research.

Strengths

• Contains 16,045 legal documents, providing a substantial corpus for legal NLP tasks.
• Includes query relevance judgments (qrels), which are essential for training and evaluating information retrieval systems.
• Focuses on a specific legal domain (Brazilian Federal Court of Accounts jurisprudence) and language (Brazilian Portuguese).

Limitations

• The dataset size of 16,045 documents may be limited for training large-scale neural retrieval models from scratch.
• Scope is restricted to jurisprudence from a single Brazilian court, which may limit generalizability to other legal systems or languages.
• Lack of publicly available column details and sample data makes preliminary data exploration and schema understanding difficult.

Provenance

Source

Jurisprudence from the Brazilian Federal Court of Accounts (Tribunal de Contas da União – TCU).

Collection Method

null

Time Range

null

Freshness

null

Geography

Brazil