vi-text-corpus / README.md
DuongTrongChi's picture
Upload dataset
04ddd68 verified
metadata
dataset_info:
  - config_name: filter-by-ppl-and-length
    features:
      - name: id
        dtype: string
      - name: revid
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
      - name: ppl
        dtype: float64
    splits:
      - name: train
        num_bytes: 69827402
        num_examples: 28269
    download_size: 123289384
    dataset_size: 69827402
  - config_name: filtered-by-rule
    features:
      - name: id
        dtype: string
      - name: revid
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
      - name: ppl
        dtype: float64
    splits:
      - name: train
        num_bytes: 480065023.4760376
        num_examples: 459402
    download_size: 555208649
    dataset_size: 480065023.4760376
  - config_name: mix-wikihow-wikipedia
    features:
      - name: text
        dtype: string
    splits:
      - name: train
        num_bytes: 2091623947
        num_examples: 549399
    download_size: 997843165
    dataset_size: 2091623947
  - config_name: ppl
    features:
      - name: id
        dtype: string
      - name: revid
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
      - name: ppl
        dtype: float64
    splits:
      - name: train
        num_bytes: 1352753394
        num_examples: 1294528
    download_size: 602025839
    dataset_size: 1352753394
  - config_name: sample
    features:
      - name: id
        dtype: string
      - name: revid
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
      - name: ppl
        dtype: float64
    splits:
      - name: train
        num_bytes: 181588953
        num_examples: 20000
    download_size: 91143497
    dataset_size: 181588953
  - config_name: tokenized_sample
    features:
      - name: input_ids
        list: int32
    splits:
      - name: train
        num_bytes: 165928500
        num_examples: 10125
    download_size: 67984707
    dataset_size: 165928500
  - config_name: wiki-wikihow-vtv
    features:
      - name: text
        dtype: string
    splits:
      - name: train
        num_bytes: 2578683736
        num_examples: 675647
    download_size: 1282890988
    dataset_size: 2578683736
  - config_name: wikihow
    features:
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
      - name: metadata
        struct:
          - name: is_expert
            dtype: bool
          - name: last_updated
            dtype: 'null'
          - name: num_votes
            dtype: 'null'
          - name: percent_helpful
            dtype: 'null'
          - name: tips
            list: 'null'
          - name: views
            dtype: 'null'
    splits:
      - name: train
        num_bytes: 1011384393
        num_examples: 89997
    download_size: 464040715
    dataset_size: 1011384393
configs:
  - config_name: filter-by-ppl-and-length
    data_files:
      - split: train
        path: filter-by-ppl-and-length/train-*
  - config_name: filtered-by-rule
    data_files:
      - split: train
        path: filtered-by-rule/train-*
  - config_name: mix-wikihow-wikipedia
    data_files:
      - split: train
        path: mix-wikihow-wikipedia/train-*
  - config_name: ppl
    data_files:
      - split: train
        path: ppl/train-*
  - config_name: sample
    data_files:
      - split: train
        path: sample/train-*
  - config_name: tokenized_sample
    data_files:
      - split: train
        path: tokenized_sample/train-*
  - config_name: wiki-wikihow-vtv
    data_files:
      - split: train
        path: wiki-wikihow-vtv/train-*
  - config_name: wikihow
    data_files:
      - split: train
        path: wikihow/train-*

Wikipedia Dataset (Tiếng Việt)

📖 Giới thiệu

Bộ dữ liệu này được tải từ Wikipedia tiếng Việt, nhằm phục vụ cho các bài toán xử lý ngôn ngữ tự nhiên (NLP) như:

  • Huấn luyện mô hình ngôn ngữ (Language Model).
  • Tiền xử lý cho bài toán dịch máy (Machine Translation).
  • Nghiên cứu về ngôn ngữ tiếng Việt.

Mô tả data

  • id: định danh của bài viết
  • revid: revision ID của Wikipedia
  • url: đường dẫn trực tiếp đến bài viết
  • title: tiêu đề bài viết
  • text: nội dung văn bản trích xuất
  • ppl: perplexity (chỉ số đánh giá chất lượng mô hình ngôn ngữ trên văn bản đó)

📂 Cấu trúc dữ liệu

  • viwiki-latest-pages-articles.xml.bz2 – dump gốc từ Wikipedia.
  • wiki_text.txt – văn bản thô đã được tách từ XML (nếu bạn đã xử lý).
  • tokenized.txt – dữ liệu đã được tokenization (nếu có).
  • vi.sp.model – mô hình SentencePiece dùng để tokenize dữ liệu (nếu có).
  • vi.arpa.bin – mô hình ngôn ngữ KenLM đã huấn luyện (nếu có).

⚙️ Nguồn dữ liệu

  • Tải trực tiếp từ Wikimedia Dumps (https://dumps.wikimedia.org/)
  • Phiên bản: viwiki-YYYYMMDD (hãy thay bằng ngày bạn tải)
  • Ngôn ngữ: Tiếng Việt

📏 Thống kê (ví dụ mẫu)

  • Số bài viết: ~X triệu.
  • Số câu: ~Y triệu.
  • Số token: ~Z triệu.