Datasets:

Modalities:
Text
Formats:
parquet
Languages:
Japanese
Libraries:
Datasets
pandas
License:
niilc-qa / README.md
teruo6939's picture
Update README.md
e628e7a verified
metadata
language:
  - ja
license: cc-by-sa-4.0
task_categories:
  - question-answering
dataset_info:
  - config_name: default
    features:
      - name: qid
        dtype: string
      - name: question
        dtype: string
      - name: answers
        sequence: string
    splits:
      - name: validation
        num_bytes: 92027
        num_examples: 795
      - name: test
        num_bytes: 23157
        num_examples: 198
    download_size: 67999
    dataset_size: 115184
  - config_name: v1.1
    features:
      - name: qid
        dtype: string
      - name: question
        dtype: string
      - name: answers
        sequence: string
      - name: D_3
        sequence: string
    splits:
      - name: test
        num_bytes: 26020
        num_examples: 198
    download_size: 16544
    dataset_size: 26020
  - config_name: v1.2
    features:
      - name: id
        dtype: string
      - name: text
        dtype: string
      - name: answers
        sequence: string
      - name: B1
        dtype: string
      - name: B2
        dtype: string
      - name: B3
        dtype: string
      - name: B4
        dtype: string
      - name: B5
        dtype: string
      - name: B6
        dtype: string
      - name: B7
        dtype: string
      - name: C_1
        dtype: string
      - name: C_2
        dtype: string
      - name: D_2
        dtype: string
      - name: D_3
        sequence: string
      - name: D_4
        dtype: string
      - name: E_1
        dtype: string
      - name: E_2
        dtype: string
      - name: E_3
        dtype: string
      - name: E_4
        dtype: string
      - name: E_5
        dtype: string
      - name: F_1
        dtype: string
      - name: F_2
        dtype: string
      - name: F_memo
        dtype: string
      - name: E_6
        dtype: string
      - name: E_7
        dtype: string
      - name: C_memo
        dtype: string
      - name: E_memo
        dtype: string
      - name: A_memo
        dtype: string
      - name: E_8
        dtype: string
      - name: D_memo
        dtype: string
      - name: A_3
        dtype: string
    splits:
      - name: test
        num_bytes: 123340
        num_examples: 198
      - name: dev
        num_bytes: 493483
        num_examples: 795
    download_size: 336210
    dataset_size: 616823
configs:
  - config_name: default
    data_files:
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
  - config_name: v1.1
    data_files:
      - split: test
        path: v1.1/test-*
  - config_name: v1.2
    data_files:
      - split: test
        path: v1.2/test-*
      - split: dev
        path: v1.2/dev-*

評価スコアの再現性確保と SB Intuitions 修正版の公開用クローン

NIILC Question Answering Dataset

This project provides a dataset for supporting the development of question answering systems. The data includes questions and their answers, and meta information such as question types, clues to obtain answers, Wikipedia pages in which answers can be found, and SPARQL queries for JWO (Japanese Wikipedia Ontology).

Licensing Information

Creative Commons Attribution Share Alike 4.0 International

Citation Information

@inproceedings{sekine2003question,
  title={Development of a question answering system focused on an encyclopedia},
  author={Sekine, Satoshi},
  booktitle={9th Annual Meeting of the Association for Natural Language Processing},
  year={2003},
  language={Japanese}
}

Subsets

default

  • qid (str): 質問を一意識別するためのID
  • question (str): 質問文
  • answers (list[str]): 質問に対する回答のリスト
    • (回答がない質問(answers: ['-'])は除外した)

v1.1

  • split: test を修正
    • NIILC-ECQA2015_test.xml から D_3列を追加
      • D_3 (str): 考えられる解答数
    • 正解文字列(answers の要素)の微修正
      • e.g., セミコロンが全角で分割できていなかった正解文字列を分割して再登録(['星条旗;古き栄光'] -> ['星条旗','古き栄光']

v1.2

両方の split を修正

  • 各質問にソースからメタデータ列をすべて追加
    • 各列の詳細はここを参照
  • split名をソースに合わせて修正
    • validation -> dev
  • 列名をソースに合わせて修正
    • qid -> id
    • question -> text
  • dev の answers に対し、v1.1 の test と同様の修正を実施