Pindorama Corpus: Portuguese Literary Texts from Brazil's Public Domain

A research-oriented Portuguese-language literary corpus derived from PDFs obtained from Brazil's Portal Domínio Público. The dataset is designed for reproducible LLM pretraining experiments, language-modeling coursework, and Portuguese NLP research. It was created by tylerxdurden and last updated on 2026-05-07.

Use Cases

• Pretraining large language models based on Portuguese literary text.
• Conducting corpus linguistics analysis on public-domain Brazilian literature.
• Teaching language-modeling concepts using a reproducible, research-oriented corpus.
• Supporting digital humanities research on Portuguese-language texts.

Strengths

• Derived from an authoritative source: Brazil's government-run Portal Domínio Público.
• Explicitly designed for reproducible research in LLM pretraining and NLP.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Data may reflect geographic and source bias inherent to the single public-domain portal.

Provenance

Source

Brazil's Portal Domínio Público (public-domain works portal).

Collection Method

Derived from PDFs obtained from the portal.

Freshness

Last updated 2026-05-07 17:45:47; freshness should be verified.

Geography

Brazil