README.md

---
configs:
- config_name: Portugal (PT)
  data_files: pt/*.parquet
- config_name: Brazil (BR)
  data_files: br/*.parquet
- config_name: Other
  data_files: other/*.parquet
task_categories:
- text-generation
- text-classification
language:
- pt
tags:
- pt
- br
- portuguese
- brazilian
- portugal
- brazil
size_categories:
- 100M<n<1B
---
# Dataset Summary
The VeraCruz Dataset is a comprehensive collection of Portuguese language content, showcasing the linguistic and cultural diversity of of Portuguese-speaking regions. It includes around 190 million samples, organized by regional origin as indicated by URL metadata into primary categories. The primary categories are:

- **Portugal (PT)**: Samples with content URLs indicating a clear Portuguese origin.
- **Brazil (BR)**: Samples with content URLs indicating a clear Brazilian origin.
- **Other**: Samples where the URL metadata does not clearly indicate a Portuguese or Brazilian origin. These samples were further classified into "PT" or "BR" categories using the [PeroVaz_PT-BR_Classifier](https://huggingface.co/Bastao/PeroVaz_PT-BR_Classifier), which is trained specifically to distinguish between the European and Brazilian variations of Portuguese.
Each entry in this category is supplemented with two extra columns: 'label' and 'score'.
The 'label' column indicates the predicted category (PT or BR), and the 'score' column represents the probability of the predicted label.

# Source Data
The VeraCruz Dataset is derived from the [MyCulturaX](https://huggingface.co/datasets/uonlp/CulturaX) dataset's Portuguese language segment, a comprehensive collection known for its broad linguistic coverage across multiple languages.
However, the original [MyCulturaX](https://huggingface.co/datasets/uonlp/CulturaX) dataset does not differentiate between the two variants of Portuguese.


# Personal and Sensitive Information
Given the dataset's extensive nature, it may contain personal and sensitive information. Users are advised to handle the data responsibly, employing ethical practices and privacy-compliant measures such as data anonymization where necessary. It is crucial to respect individual privacy and adhere to legal standards when utilizing this dataset.

# Licensing Information
The license terms for the VeraCruz Dataset strictly follow those of mC4 and OSCAR. Please refer to the licenses of both datasets when using VeraCruz:

- [mC4 License Details](https://huggingface.co/datasets/allenai/c4#license)
- [OSCAR License Details](https://huggingface.co/datasets/oscar-corpus/OSCAR-2301#licensing-information)