Hugging Face's logo

Datasets:
nampdn-ai
/
vietvault

Tasks:
Text Generation
Modalities:
Tabular
Text
Formats:
parquet
Languages:
Vietnamese
Size:
1M - 10M
DOI:
Libraries:
Datasets
Dask
License:
Dataset card Viewer Files Community
1
vietvault / README.md
nampdn-ai's picture
nampdn-ai
Update README.md
99864a4 verified
preview code
|
raw
history blame contribute delete
5.57 kB
metadata
license: odc-by
extra_gated_prompt: >
  # VietVault Data Usage Agreement


  By using the VietVault dataset, you agree to the following terms:


  1. **License**: Use the dataset under the ODC-By license.

  2. **Attribution**: Cite the VietVault dataset card and acknowledge the Common
  Crawl dataset.

  3. **Non-Commercial Use**: Obtain written permission for commercial use.

  4. **No Warranty**: The dataset is provided "as is" without warranty.

  5. **Limitation of Liability**: Dataset owners are not liable for any damages.

  6. **Ethical Use**: Use the dataset responsibly and ethically, complying with
  laws and regulations.

  7. **Redistribution**: Include this agreement and the ODC-By license when
  redistributing.

  8. **Termination**: Dataset owners may terminate access if terms are breached.

  9. **Governing Law**: Agreement governed by the laws of [Jurisdiction].


  By using the VietVault dataset, you acknowledge and agree to these terms.
extra_gated_fields:
  Name: text
  Email: text
  Affiliation: text
  Country: text
  I agree to the LICENSE of this dataset: checkbox
task_categories:
  - text-generation
language:
  - vi
pretty_name: Vietnamese Common Crawl
size_categories:
  - 10M<n<100M

VietVault

VietVault is a large-scale Vietnamese language corpus, carefully filtered and curated from Common Crawl dataset dumps prior to 2023. This dataset is designed to serve as a high-quality resource for Vietnamese language model pretraining and various natural language processing tasks.

Dataset Statistics

  • Size: 80GB of raw text
  • Language: Vietnamese
  • Source: Common Crawl dataset (all dumps in 2013-2023)
  • Preprocessing: Cleaned, deduplicated, filtered for Vietnamese language content, and toxic content removed

Data Filtering and Cleaning

The raw data was obtained from the Common Crawl dataset, which consists of web page content collected from various sources across the internet. To create VietVault, the following steps were performed:

  1. Language Identification: Vietnamese language content was identified and selected from the Common Crawl dataset using language detection techniques.
  2. Deduplication: Duplicate content was removed to ensure the dataset consists of unique Vietnamese text.
  3. Cleaning: The filtered data was cleaned by removing noise, irrelevant characters, and formatting issues to enhance the quality of the corpus.
  4. Toxic Content Filtering: The dataset underwent an additional filtering step to remove toxic, offensive, and inappropriate content, ensuring a safer and more ethical resource for training language models.

Intended Use and Applications

VietVault is primarily intended to be used as a pretraining dataset for Vietnamese language models. It can be utilized for tasks such as:

  • Unsupervised pretraining of Vietnamese language models (e.g., BERT, RoBERTa, GPT)
  • Fine-tuning pretrained models for downstream Vietnamese NLP tasks (e.g., text classification, named entity recognition, question answering)
  • Linguistic analysis and research on Vietnamese language characteristics and patterns

Data Format

The dataset is provided in plain Markdown format converted from HTML using trafilatura, with each line representing a separate document or text sample from the internet. The text is encoded in UTF-8 to ensure compatibility with various processing tools and frameworks.

License and Attribution

VietVault is licensed under the Open Data Commons Attribution License (ODC-By). By using this dataset, you agree to the terms and conditions of the ODC-By license. You are free to share, create, and adapt the dataset as long as you attribute VietVault and the original source (Common Crawl dataset). Please see the full text of the ODC-By license for more details: https://opendatacommons.org/licenses/by/1-0/

Acknowledgement

I would like to express my deepest gratitude and utmost respect to the following exceptional individuals for their invaluable contributions to this project:

tiendung@symato, que-nguyen@symato, binhvq, th1nhng0, and iambestfeed have demonstrated remarkable expertise and unwavering dedication in developing the Python script that allowed for the efficient download and filtering of the extensive Common Crawl (CC) dumps. Their efforts were crucial in extracting the Vietnamese corpus from the raw data, a task that required immense skill and meticulous attention to detail.

Furthermore, I am profoundly grateful for their generous contribution of computing resources, including CPUs, network bandwidth, and storage capacity. Processing the staggering 6TB of raw, unclean Vietnamese text demanded significant computational power, and these individuals selflessly provided the necessary infrastructure to accomplish this challenging task.

Their exceptional commitment, technical prowess, and unwavering support have been truly awe-inspiring. I am incredibly fortunate to have benefited from their expertise and generosity in advancing the field of natural language processing for the Vietnamese language.

Contact and Feedback

If you have any questions, feedback, or suggestions regarding VietVault, please create an issue in this repo.

I hope that VietVault serves as a valuable resource for advancing Vietnamese language processing and understanding while promoting responsible and ethical language model development. Happy exploring and modeling!