ja-stackoverflow / README.md
p1atdev's picture
Update README.md
f1cbe26
metadata
dataset_info:
  - config_name: default
    features:
      - name: question
        struct:
          - name: accepted_answer_id
            dtype: string
          - name: answer_count
            dtype: int64
          - name: body
            dtype: string
          - name: comment_count
            dtype: int64
          - name: content_license
            dtype: string
          - name: creation_date
            dtype: string
          - name: favorite_count
            dtype: int64
          - name: id
            dtype: string
          - name: last_activity_date
            dtype: string
          - name: last_edit_date
            dtype: string
          - name: last_editor_user_id
            dtype: string
          - name: owner_user_id
            dtype: string
          - name: post_type
            dtype: string
          - name: score
            dtype: int64
          - name: tags
            sequence: string
          - name: title
            dtype: string
          - name: view_count
            dtype: int64
      - name: answers
        list:
          - name: body
            dtype: string
          - name: comment_count
            dtype: int64
          - name: content_license
            dtype: string
          - name: creation_date
            dtype: string
          - name: id
            dtype: string
          - name: last_activity_date
            dtype: string
          - name: last_edit_date
            dtype: string
          - name: last_editor_user_id
            dtype: string
          - name: owner_user_id
            dtype: string
          - name: parent_id
            dtype: string
          - name: post_type
            dtype: string
          - name: score
            dtype: int64
      - name: id
        dtype: string
      - name: accepted_answer_id
        dtype: string
      - name: popular_answer_id
        dtype: string
    splits:
      - name: train
        num_bytes: 112596554
        num_examples: 30551
    download_size: 54805530
    dataset_size: 112596554
  - config_name: simple
    features:
      - name: id
        dtype: string
      - name: accepted_answer_id
        dtype: string
      - name: popular_answer_id
        dtype: string
      - name: title
        dtype: string
      - name: question_body
        dtype: string
      - name: question_score
        dtype: int64
      - name: accepted_answer_body
        dtype: string
      - name: accepted_answer_score
        dtype: int64
      - name: popular_answer_body
        dtype: string
      - name: popular_answer_score
        dtype: int64
      - name: tags
        sequence: string
    splits:
      - name: train
        num_bytes: 113051344
        num_examples: 30551
    download_size: 56632072
    dataset_size: 113051344
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
  - config_name: simple
    data_files:
      - split: train
        path: simple/train-*
license: cc-by-sa-4.0
task_categories:
  - text-generation
  - question-answering
language:
  - ja
tags:
  - stackoverflow
  - programming
pretty_name: Japanese StackOverflow
size_categories:
  - 10K<n<100K

ja-stackoverflow

日本語版 Stack Overflow の スタック・オーバーフローデータダンプ をもとにデータを加工し、質問文と回答文のペアになるように調整した QA データセット。

データ構造

投稿本文は html2text を使ってマークダウン化されています。その際、

  • コードブロックは ``` で囲まれるように変更されています。
  • 画像 URL に base64 エンコードされた画像が含まれる場合、 [unk] に置き換えています。

default サブセット

  • id: 質問投稿の ID
  • question: 質問投稿
  • answers: 質問に対する回答投稿のリスト
  • accepted_answer_id: 質問者に選ばれた回答のID。null の可能性がある
  • popular_answer_id: もっともスコアが高かった回答のID。null の可能性がある

simple サブセット

default サブセットから、 questionanswers の辞書を展開しシンプルにしたもの。

  • id: 質問投稿の ID
  • accepted_answer_id: 質問者に選ばれた回答のID。null の可能性がある
  • popular_answer_id: もっともスコアが高かった回答のID。null の可能性がある
  • title: 質問のタイトル
  • question_body: 質問本文
  • question_score: 質問のスコア
  • tags: 質問に関連付けられたタグ
  • accepted_answer_body: 質問者に選ばれた回答の本文。null の可能性がある
  • accepted_answer_score: 質問者に選ばれた回答のスコア。null の可能性がある
  • popular_answer_body: もっともスコアが高かった回答の本文。null の可能性がある
  • popular_answer_score: もっともスコアが高かった回答のスコア。null の可能性がある

使い方

datasets ライブラリを用いて簡単に利用できます。

from datasets import load_dataset

dataset = load_dataset("p1atdev/ja-stackoverflow", name="simple" split="train")

print(dataset)
#Dataset({
#    features: ['id', 'accepted_answer_id', 'popular_answer_id', 'title', 'question_body', 'question_score', 'accepted_answer_body', 'accepted_answer_score', 'popular_answer_body', 'popular_answer_score', 'tags'],
#    num_rows: 30551
#})

ライセンス

StackOverflow に基づき、CC BY-SA 4.0