Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

shivangibithel/SOTAB(複製)

データの内容

半構造化された表の列に Schema.org の型を付けた、列型アノテーションのデータ。1 行が 1 つの列で、column_index に列の位置、label に型名(例は currency)、table に表の名前・ヘッダー・行が入る。ラベルは Web ページの Schema.org マークアップに由来し、サイト側の付け間違いを含む(調査メモによる)。ライセンスは CC BY-SA 4.0 とカードにある。

データの概要

項目 値
行数 162,351(1 構成・3 分割の合計)
列数 3
容量 21.43 KB(ファイル 3 本)
ファイルの形式 py 1
言語 en
公開 public

構成と分割

構成(config) 分割(split)ごとの行数 列数
default train 130,471、test 15,040、validation 16,840 3

列(構成 default)

列 型
column_index int32
label string
table struct{header, rows, name}

行数と列は元のカードの YAML(dataset_info)の記載。データセットビューアでは読めなかった。全データセットの一覧は HYUKI5/dataset-catalog にある。

複製について

shivangibithel/SOTAB をリビジョン a90d318e8c5a3333038c9e6ce4e3cbf21857cf38 の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。

出典

論文

アノテーションの方法

データセットの作成者

Panupong Pasupat and Percy Liang

微調整に使うときの注意

  • 調査メモの台帳での区分: B(継承条件・gated・モデル由来の条件・上流の規約などが付く)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
  • Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
  • 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。

調査メモの記述

メモ 節 ライセンス(記載) 確認 注記
ontology_tables_schema_kg.md Q1. 表の列・セル・列間関係に、意味型やエンティティのラベルが付いたデータは何があるか (L26) ページにライセンス記載なし。HF ミラーは cc-by-4.0 のタグだが、カード本文は別データ(WikiTableQuestions)の記述が混ざり(Cura 確認済 複数/混在 / 原文注記: ミラーは SOTAB.py のスクリプトのみで、datasets 4 以降では読めない可能性(未検証)。ラベルは Web の Schema.org マークアップ由来で、サイト側の付け間違いを含む(V2 で test を人

元のカード(日本語訳)

以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。


SOTAB のデータセットカード

目次

  • データセットの説明
    • データセットの概要
    • 対応するタスク
    • 言語
  • データセットの構造
    • データのインスタンス
    • データのフィールド
    • データの split
  • データセットの作成
    • 収集の根拠
    • 元データ
    • アノテーション
    • 個人情報と機微な情報
  • データ利用上の留意点
    • データセットの社会的影響
    • バイアスについての議論
    • そのほかの既知の制約
  • 補足情報
    • データセットの作成者
    • ライセンス情報
    • 引用情報

データセットの説明

データセットの概要

SOTAB データセットは、半構造化テーブルにおける列型アノテーションのタスクのための大規模なデータセットである。

対応するタスクとリーダーボード

table-annotation, column-type-annotation

言語

英語 (en)

データセットの構造

データのインスタンス

default

  • ダウンロードしたデータセットファイルのサイズ:
  • 生成されたデータセットのサイズ:
  • ディスクの総使用量:

'validation' の例は次のとおり:

{
    "id": 0,
    "column_index": 3,
    "label": "currency",
    "table": {
        "name": "Book_7sat.co.uk_September2020_CTA.json.gz",
        "header": ["col1", "col2", ...]
        "rows": [
           ["2001", "2", "USL A-League", ...],
           ["2002", "2", "USL A-League", ...], 
           ...
        ]
    }
}

データのフィールド

データのフィールドは、すべての split で同じである。

default

  • id: int32 の特徴量。
  • column_index: string の特徴量。
  • label: string の特徴量。
  • table: 次を含む辞書の特徴量:
    • rows: string の特徴量の list。
    • rows: string の特徴量の list の list。
    • name: string の特徴量。

データの split

name train validation test
default 130471 16840 15040

データセットの作成

収集の根拠

[Needs More Information]

元データ

初期のデータ収集と正規化

[Needs More Information]

元の言語データの作成者は誰か?

[Needs More Information]

アノテーション

アノテーションの過程

[Needs More Information]

アノテーターは誰か?

[Needs More Information]

個人情報と機微な情報

[Needs More Information]

データ利用上の留意点

データセットの社会的影響

[Needs More Information]

バイアスについての議論

[Needs More Information]

そのほかの既知の制約

[Needs More Information]

補足情報

データセットの作成者

Panupong Pasupat and Percy Liang

ライセンス情報

Creative Commons Attribution Share Alike 4.0 International

引用情報


貢献

このデータセットの追加に協力した @ShivangiBithel に感謝する。

原文(英語)

Dataset Card for SOTAB

Dataset Summary

The SOTAB dataset is a large-scale dataset for the task of column type annotation on semi-structured tables.

Supported Tasks and Leaderboards

table-annotation, column-type-annotation

Languages

en

Dataset Structure

Data Instances

default

  • Size of downloaded dataset files:
  • Size of the generated dataset:
  • Total amount of disk used:

An example of 'validation' looks as follows:

{
    "id": 0,
    "column_index": 3,
    "label": "currency",
    "table": {
        "name": "Book_7sat.co.uk_September2020_CTA.json.gz",
        "header": ["col1", "col2", ...]
        "rows": [
           ["2001", "2", "USL A-League", ...],
           ["2002", "2", "USL A-League", ...], 
           ...
        ]
    }
}

Data Fields

The data fields are the same among all splits.

default

  • id: a int32 feature.
  • column_index: a string feature.
  • label: a string feature.
  • table: a dictionary feature containing:
    • rows: a list of string features.
    • rows: a list of list of string features.
    • name: a string feature.

Data Splits

name train validation test
default 130471 16840 15040

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Who are the source language producers?

[Needs More Information]

Annotations

Annotation process

[Needs More Information]

Who are the annotators?

[Needs More Information]

Personal and Sensitive Information

[Needs More Information]

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

[Needs More Information]

Additional Information

Dataset Curators

Panupong Pasupat and Percy Liang

Licensing Information

Creative Commons Attribution Share Alike 4.0 International

Citation Information


Contributions

Thanks to @ShivangiBithel for adding this dataset.

Downloads last month
91