---
license: apache-2.0
dataset_info:
  features:
  - name: 'Unnamed: 0'
    dtype: int64
  - name: text
    dtype: string
  - name: timestamp
    dtype: string
  - name: url
    dtype: string
  - name: label
    dtype: int64
  splits:
  - name: train
    num_bytes: 3510227581
    num_examples: 836407
  download_size: 1801797299
  dataset_size: 3510227581
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
Data sources come from the following categories:
1.Web crawler dataset:  
- Website UET (ĐH Công nghệ): tuyensinh.uet.vnu.edu.vn; new.uet.vnu.edu.vn
- Website HUS (ĐH KHTN): hus.vnu.edu.vn
- Website EUB (ĐH Kinh tế): ueb.vnu.edu.vn
- Website IS (ĐH Quốc tế): is.vnu.edu.vn
- Website Eduacation (ĐH Giáo dục): education.vnu.edu.vn
- Website NXB ĐHQG: press.vnu.edu.vn   
[List domain web crawler](https://docs.google.com/spreadsheets/d/1zbkltkSPRm6f48Lb1Jo3Njq1-LrSd8H6/edit?gid=409337688#gid=409337688)
2. CC100:  
[link to CC100 vi](https://huggingface.co/datasets/statmt/cc100)  
3. Vietnews:  
  [link to bk vietnews dataset](https://github.com/ThanhChinhBK/vietnews)
4. C4_vi:  
  [link to C4_vi](https://huggingface.co/datasets/allenai/c4)
  
Folder Toxic store files demo for toxic filtering.We filtered C4_validation dataset, vietnews samples dataset and a part(1/50) of CC100_vi dataset.After the process, datasets split into nontoxic part and toxic part.

Folder Dedup store files after the deduplication process of above files.  
Folder Toxic_2, Dedup_2, Tokenized_2 is the result of the second process that we executed on 17 files of C4_vi dataset wich contains 1b tokens