|
--- |
|
license: cc |
|
language: |
|
- eu |
|
tags: |
|
- Basque |
|
- Euskara |
|
pretty_name: ZelaiHandi🌱 |
|
version: v1 |
|
release_date: 2024/03/04 |
|
size_categories: |
|
- 100M<n<1B |
|
--- |
|
<!--# ZelaiHandi🌱: A Large Collection of Basque texts--> |
|
|
|
<img src="https://storage.googleapis.com/orai-nlp/datasets/ZelaiHandi/zelaihandi_large.png" alt="ZelaiHandi_logo" width=800 /> |
|
|
|
# Dataset Card for ZelaiHandi🌱 |
|
|
|
<!-- Provide a quick summary of the dataset. --> |
|
ZelaiHandi🌱: A Large Collection of Basque Texts. |
|
ZelaiHandi, which means "large pasture" in Basque, is the largest collection of freely licensed and clean Basque texts to date (March 4th, 2024), gathered from selected web sources. |
|
This collection comprises approximately 521 million tokens. |
|
|
|
The dataset will receive periodical updates. |
|
|
|
The corpus has been released with the objective of “feeding” Large Language Models. Naturally, models that are not strictly "herbivorous" can also ingest ZelaiHandi. |
|
|
|
## Dataset Details |
|
|
|
### Dataset Description |
|
|
|
<!-- Provide a longer summary of what this dataset is. --> |
|
|
|
- **Size:** 521M words (4.5GB disk space) |
|
- **Language:** Basque (EU) |
|
- **License:** Various CC licenses |
|
- **Curated by:** Orai NLP Teknologiak |
|
|
|
### Dataset Structure |
|
|
|
<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. --> |
|
|
|
The corpus is released in JSONL format. Each document contains the following attributes: |
|
|
|
```json |
|
{ |
|
"source": "source", |
|
"license": "document license", |
|
"lang": "eu", |
|
"url": "document url", |
|
"title": "document title", |
|
"author": "document author if known, null otherwise", |
|
"date": "publication date in yyyy[-mm-dd[Thh:mm:ssZ]] format", |
|
"text": "document content, ready to use raw text. Title text (for news and articles) is included as well.", |
|
"domain": "document domain. one of ['news', 'administrative', 'wikipedia', 'science', 'subtitles', 'literature', 'videogames']" |
|
} |
|
``` |
|
|
|
|
|
<!--## Dataset Creation--> |
|
|
|
### Source Data |
|
|
|
<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). --> |
|
This is a corpus of the Basque language. It contains 521,546,047 raw words (counted by the '```wc -w```' command over the text attribute) in 1.63M documents. It was gathered from various web sources and contains data up to 2024/02/19. |
|
|
|
The following table summarizes corpus statistics and license information. |
|
|
|
| **source** | **base url** | **domain** | **tokens (M)** | **license** | |
|
|----------------------------:|------------------------------------------------------------------------------------------------:|---------------:|---------------:|-----------------------------:| |
|
| Tokikom 📝 | https://tokikom.eus/bazkideak/ | news | 163.72 | cc-by / cc-by-sa | |
|
| Berria | https://berria.eus | news | 125.72 | cc-by-sa 4.0 | |
|
| Eusko Legebiltzarra | https://www.legebiltzarra.eus | administrative | 80.16 | public domain | |
|
| Wikipedia | https://wikipedia.org | wikipedia | 63.83 | cc-by-sa 4.0 | |
|
| Argia | https://www.argia.eus | news | 17.44 | cc-by-sa 4.0 | |
|
| Addi | https://addi.ehu.es/ | science | 17.09 | several cc variants | |
|
| Opendata Euskadi subtiitles | https://opendata.euskadi.eus/catalogo/-/euskarazko-azpitituluak-filmak-eta-telesailak-euskaraz/ | subtitles | 11.30 | cc-by-sa | |
|
| Hitza | https://hitza.eus | news | 9.57 | cc-by-sa 4.0 | |
|
| GFA | https://www.bngipuzkoa.eus/ | administrative | 8.46 | Custom copyright license⚠️ | |
|
| Zientzia.eus | https://zientzia.eus | science | 8.37 | cc-by-sa | |
|
| Susa | https://www.susa-literatura.eus/ | literature | 5.77 | cc-by | |
|
| BFA | https://jjggbizkaia.eus | administrative | 4.26 | Custom copyright license⚠️ | |
|
| Zientzia Kaiera | https://zientziakaiera.eus/ | science | 1.94 | cc-by-sa 3.0 | |
|
| Ekaia | https://ojs.ehu.eus/index.php/ekaia | science | 1.80 | cc-by-nc-nd 4.0 | |
|
| Ikergazte | https://www.buruxkak.eus/bilatu?bilaketa=ikergazte | science | 1.68 | cc-by-sa 3.0 | |
|
| Game-erauntsia | https://gamerauntsia.eus/ | videogames | 0.40 | cc-by-sa 4.0 | |
|
| **Total** | | | **521.55** | |
|
|
|
📝 Tokikom is a Basque local media network. The following websites are included in ZelaiHandi: [28kanala, Aiaraldea, Aikor, Aiurri, Alea, Amezti, Anboto, Antxetamedia, Ataria, Baleike, Barren, Begitu, Berriketan, Drogetenitturri, Erran, Euskalerriairratia, Gitb, Goiena, Guaixe, Hiruka, Karkara, Kronika, Mailope, Maxixatzen, Noaua, Plaentxia, Txintxarri, Uriola, Urolakostakohitza, Uztarria, Zarauzkohitza]. |
|
|
|
⚠️ Custom license, allows reproducibility. License text here: <https://www.bngipuzkoa.eus/WAS/CORP/DJGPortalWEB/aviso_legal.jsp>. |
|
|
|
⚠️ Custom license, allows reproducibility License text here: <https://jjggbizkaia.eus/eu/pribadutasun-politikea>. |
|
|
|
<img src="https://storage.googleapis.com/orai-nlp/datasets/ZelaiHandi/zelaihandi_domainstats.png" alt="domain distribution statistics of ZelaiHandi corpus" width=800 /> |
|
|
|
|
|
## License |
|
|
|
The corpus is comprised of documents from various sources. All documents included have a free license, but each source has its own license. All documents collected are publicly available under free use licenses in their original website. If you are the owner of any data included in the corpus and you would not like to be reproduced here, please contact us. |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
<!-- This section is meant to convey both technical and sociotechnical limitations. --> |
|
This corpus may reflect societal stereotypes and biases. Additionally, it could contain explicit language, slurs, and hate speech, particularly in fictional texts from literature. |
|
|
|
<!--### Recommendations --> |
|
|
|
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> |
|
|
|
<!--Users should be made aware of the risks, biases and limitations of the dataset. |
|
We suggest filtering sources as literature and subtitles that might contain hate speech, slurs or explicit language before pretraining an LLM for end tasks like interactive chat.--> |
|
|
|
## Citation |
|
|
|
<!--We plan to release a paper soon, but, for now, please, cite as follows:--> |
|
If you use the corpus please cite the following reference: |
|
|
|
<!--**BibTeX:**--> |
|
```bibtex |
|
@misc{ZelaiHandi, |
|
title = {ZelaiHandi: A Large Collection of Basque Texts}, |
|
author = {Iñaki San Vicente, Gorka Urbizu, Ander Corral, Zuhaitz Beloki, Xabier Saralegi}, |
|
publisher = {Orai NLP Technologies}, |
|
url = {\url{https://huggingface.co/datasets/orai-nlp/ZelaiHandi}}, |
|
year = 2024 } |
|
``` |
|
|
|
## Acknowledgements |
|
|
|
Special thanks to the authors of the texts and to the media that have made it possible to publish their content under a free license. |
|
|
|
This work has been partially funded by the Basque Government under grant no. KK-2023/00094 (ICL4LANG). |
|
|
|
|
|
|
|
## Contact |
|
|
|
- Iñaki San Vicente (i.sanvicente@orai.eus) |
|
- Gorka Urbizu (g.urbizu@orai.eus) |
|
|
