Fineweb2 Bagaco: Portuguese Language Pretraining Corpus

A pretraining dataset for European Portuguese created by filtering the Fineweb2 dataset to URLs from Portuguese domains. Each document is classified into one of 9 categories and scored for educational quality. The dataset was created by duarteocarmo and was last updated on 2026-02-22.

Use Cases

• Pretraining large language models based on European Portuguese web content.
• Evaluating text quality and educational value based on the provided scoring.
• Analyzing content distribution across the 9 document categories.
• Comparing language model performance on domain-specific Portuguese text.

Strengths

• Documents are classified into one of 9 categories, providing structured metadata.
• Each document is scored for educational quality, enabling quality-based filtering.
• Dataset is derived from the established Fineweb2 dataset, suggesting a known base source.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

HuggingFaceFW/fineweb-2, subset por_Latn, split train

Collection Method

Filtered to URLs containing .pt/ (Portuguese top-level domain).

Freshness

Last updated 2026-02-22 13:48:25; freshness should be verified.

Geography

Likely contains content from Portuguese domains (.pt).