# Open Science Dataset
## Overview
Open Science is a large-scale, permissively licensed text dataset derived from OpenAlex, containing over 100B (105,390,332,599) words. OpenAlex is an open database of scholarly publications, authors, institutions, and research outputs that serves as a comprehensive source for academic literature.
## Key Features
- **Truly Open**: Contains only permissively licensed data suitable for both commercial and non-commercial use
- **Multilingual Coverage**: Predominantly English with significant representation of Portuguese, German, Indonesian, Russian, and Spanish
- **Academic Diversity**: Encompasses journal articles, books, datasets, theses, and other document types
## Dataset Statistics

License distribution:
```
cc-by      ███████████████████████████████ 93.09%
cc-by-sa   ██ 3.68%
public     ██ 3.23%
```
![License Distribution](license_graph.png)
- Total number of documents:
  11,534,164
  ## Language Distribution

| Language     | Document Count | Word Count      | % of Total Words |
|-------------|----------------|-----------------|------------------|
| English     | 9,692,480      | 91,961,796,098  | 87.28%          |
| Portuguese  | 381,719        | 2,911,255,370   | 2.76%           |
| German      | 167,130        | 1,606,879,128   | 1.53%           |
| Indonesian  | 297,645        | 1,370,676,366   | 1.30%           |
| Russian     | 199,317        | 1,235,729,351   | 1.17%           |
| Spanish     | 189,305        | 1,548,199,226   | 1.47%           |

## Dataset Structure
### Data Fields
| Field               |
|---------------------|
| identifier          |
| pdf_url             |
| lang                |
| error               |
| title               |
| source_name         |
| publication_year    |
| license             |
| word_count          |
| text                |

## How to Use
### Installation
```python
from datasets import load_dataset
dataset = load_dataset('PleIAs/openscource')
```
## Considerations and Limitations
- While permissively licensed, users should verify compatibility with their specific use case
- Language distribution may not be uniform
- Historical texts may contain dated terminology or perspectives
## Acknowledgements
This dataset was developed with support from:
- Jean Zay (Eviden, Idris)
- Nebius AI
- Tracto AI