HYUKI5's picture
カードを組み立て直す(訳文と書式の修正を反映)
d730caf verified
|
Raw History Blame Contribute Delete
13.1 kB
---
license: other
pretty_name: lighteval/EntityMatching (mirror)
source_datasets:
- lighteval/EntityMatching
tags:
- laya
- laya-finetune
- mirror
- domain-business
- use-unclear
- genre-table
- label-binary
- annot-unknown
- lang-en
configs:
- config_name: Abt_Buy
data_files:
- split: train
path: Abt_Buy/train-*
- split: validation
path: Abt_Buy/validation-*
- split: test
path: Abt_Buy/test-*
- config_name: Amazon_Google
data_files:
- split: train
path: Amazon_Google/train-*
- split: validation
path: Amazon_Google/validation-*
- split: test
path: Amazon_Google/test-*
- config_name: Beer
data_files:
- split: train
path: Beer/train-*
- split: validation
path: Beer/validation-*
- split: test
path: Beer/test-*
- config_name: Company
data_files:
- split: train
path: Company/train-*
- split: validation
path: Company/validation-*
- split: test
path: Company/test-*
- config_name: DBLP_ACM
data_files:
- split: train
path: DBLP_ACM/train-*
- split: validation
path: DBLP_ACM/validation-*
- split: test
path: DBLP_ACM/test-*
- config_name: DBLP_GoogleScholar
data_files:
- split: train
path: DBLP_GoogleScholar/train-*
- split: validation
path: DBLP_GoogleScholar/validation-*
- split: test
path: DBLP_GoogleScholar/test-*
- config_name: Dirty_DBLP_ACM
data_files:
- split: train
path: Dirty_DBLP_ACM/train-*
- split: validation
path: Dirty_DBLP_ACM/validation-*
- split: test
path: Dirty_DBLP_ACM/test-*
- config_name: Dirty_DBLP_GoogleScholar
data_files:
- split: train
path: Dirty_DBLP_GoogleScholar/train-*
- split: validation
path: Dirty_DBLP_GoogleScholar/validation-*
- split: test
path: Dirty_DBLP_GoogleScholar/test-*
- config_name: Dirty_Walmart_Amazon
data_files:
- split: train
path: Dirty_Walmart_Amazon/train-*
- split: validation
path: Dirty_Walmart_Amazon/validation-*
- split: test
path: Dirty_Walmart_Amazon/test-*
- config_name: Dirty_iTunes_Amazon
data_files:
- split: train
path: Dirty_iTunes_Amazon/train-*
- split: validation
path: Dirty_iTunes_Amazon/validation-*
- split: test
path: Dirty_iTunes_Amazon/test-*
- config_name: Fodors_Zagats
data_files:
- split: train
path: Fodors_Zagats/train-*
- split: validation
path: Fodors_Zagats/validation-*
- split: test
path: Fodors_Zagats/test-*
- config_name: Walmart_Amazon
data_files:
- split: train
path: Walmart_Amazon/train-*
- split: validation
path: Walmart_Amazon/validation-*
- split: test
path: Walmart_Amazon/test-*
- config_name: iTunes_Amazon
data_files:
- split: train
path: iTunes_Amazon/train-*
- split: validation
path: iTunes_Amazon/validation-*
- split: test
path: iTunes_Amazon/test-*
dataset_info:
- config_name: Abt_Buy
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 580495
num_examples: 1232
- name: validation
num_bytes: 926013
num_examples: 1916
- name: test
num_bytes: 937812
num_examples: 1916
download_size: 508497
dataset_size: 2444320
- config_name: Amazon_Google
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 262631
num_examples: 1398
- name: validation
num_bytes: 430776
num_examples: 2293
- name: test
num_bytes: 429953
num_examples: 2293
download_size: 247843
dataset_size: 1123360
- config_name: Beer
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 20547
num_examples: 80
- name: validation
num_bytes: 24191
num_examples: 91
- name: test
num_bytes: 24255
num_examples: 91
download_size: 29542
dataset_size: 68993
- config_name: Company
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 438041265
num_examples: 33718
- name: validation
num_bytes: 322303999
num_examples: 22533
- name: test
num_bytes: 309899980
num_examples: 22503
download_size: 365685795
dataset_size: 1070245244
- config_name: DBLP_ACM
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 917934
num_examples: 2664
- name: validation
num_bytes: 848473
num_examples: 2473
- name: test
num_bytes: 843238
num_examples: 2473
download_size: 824232
dataset_size: 2609645
- config_name: DBLP_GoogleScholar
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 1911584
num_examples: 6414
- name: validation
num_bytes: 1658208
num_examples: 5742
- name: test
num_bytes: 1650462
num_examples: 5742
download_size: 1611925
dataset_size: 5220254
- config_name: Dirty_DBLP_ACM
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 960882
num_examples: 2664
- name: validation
num_bytes: 882597
num_examples: 2473
- name: test
num_bytes: 877988
num_examples: 2473
download_size: 871768
dataset_size: 2721467
- config_name: Dirty_DBLP_GoogleScholar
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 1980418
num_examples: 6414
- name: validation
num_bytes: 1721356
num_examples: 5742
- name: test
num_bytes: 1713932
num_examples: 5742
download_size: 1681464
dataset_size: 5415706
- config_name: Dirty_Walmart_Amazon
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 357946
num_examples: 1152
- name: validation
num_bytes: 645692
num_examples: 2049
- name: test
num_bytes: 645688
num_examples: 2049
download_size: 487611
dataset_size: 1649326
- config_name: Dirty_iTunes_Amazon
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 93461
num_examples: 156
- name: validation
num_bytes: 66813
num_examples: 109
- name: test
num_bytes: 67621
num_examples: 109
download_size: 82094
dataset_size: 227895
- config_name: Fodors_Zagats
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 34487
num_examples: 132
- name: validation
num_bytes: 49533
num_examples: 190
- name: test
num_bytes: 49544
num_examples: 189
download_size: 42929
dataset_size: 133564
- config_name: Walmart_Amazon
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 341160
num_examples: 1152
- name: validation
num_bytes: 616628
num_examples: 2049
- name: test
num_bytes: 615948
num_examples: 2049
download_size: 460436
dataset_size: 1573736
- config_name: iTunes_Amazon
features:
- name: productA
dtype: string
- name: productB
dtype: string
- name: same
dtype: bool_
splits:
- name: train
num_bytes: 88464
num_examples: 156
- name: validation
num_bytes: 63901
num_examples: 109
- name: test
num_bytes: 64425
num_examples: 109
download_size: 70776
dataset_size: 216790
---
# lighteval/EntityMatching(複製)
## データの内容
2 つのレコードが同じものを指すかを判定するためのデータ。1 件は `productA` と `productB` の 2 つのレコードの組で、同じかどうかを `same` が示し、Abt_Buy、Beer、Company など 13 の構成に分かれる。元のカードは空で、この説明は列名と調査メモから書いた。
## データの概要
| 項目 | 値 |
|---|---|
| 行数 | 152,839(13 構成・39 分割の合計) |
| 列数 | 3 |
| 容量 | 372.61 MB(ファイル 41 本) |
| ファイルの形式 | parquet 39 |
| 公開 | public |
### 構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---:|
| `Abt_Buy` | `train` 1,232、`validation` 1,916、`test` 1,916 | 3 |
| `Amazon_Google` | `train` 1,398、`validation` 2,293、`test` 2,293 | 3 |
| `Beer` | `train` 80、`validation` 91、`test` 91 | 3 |
| `Company` | `train` 33,718、`validation` 22,533、`test` 22,503 | 3 |
| `DBLP_ACM` | `train` 2,664、`validation` 2,473、`test` 2,473 | 3 |
| `DBLP_GoogleScholar` | `train` 6,414、`validation` 5,742、`test` 5,742 | 3 |
| `Dirty_DBLP_ACM` | `train` 2,664、`validation` 2,473、`test` 2,473 | 3 |
| `Dirty_DBLP_GoogleScholar` | `train` 6,414、`validation` 5,742、`test` 5,742 | 3 |
| `Dirty_Walmart_Amazon` | `train` 1,152、`validation` 2,049、`test` 2,049 | 3 |
| `Dirty_iTunes_Amazon` | `train` 156、`validation` 109、`test` 109 | 3 |
| `Fodors_Zagats` | `train` 132、`validation` 190、`test` 189 | 3 |
| `Walmart_Amazon` | `train` 1,152、`validation` 2,049、`test` 2,049 | 3 |
| `iTunes_Amazon` | `train` 156、`validation` 109、`test` 109 | 3 |
### 列(構成 `Abt_Buy`・分割 `train`。どの構成も同じ列)
| 列 | 型 | 先頭の行の値 |
|---|---|---|
| `productA` | `string` | `name: canon high capacity color ink cartridge co…` |
| `productB` | `string` | `name: canon cli-221 black ink cartridge 2946b001…` |
| `same` | `bool_` | `False` |
行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は [HYUKI5/dataset-catalog](https://huggingface.co/datasets/HYUKI5/dataset-catalog) にある。
## 複製について
[`lighteval/EntityMatching`](https://huggingface.co/datasets/lighteval/EntityMatching) をリビジョン `dbc9ec9d4de43d805215f0f7da051673bd938747` 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
- 元のデータセット: https://huggingface.co/datasets/lighteval/EntityMatching
- 複製したリビジョン: `dbc9ec9d4de43d805215f0f7da051673bd938747`(2025-08-18 13:15:21+00:00)
- 元のライセンス表記: `other`
## 論文
- 元のカードに論文へのリンクは見つからなかった
## アノテーションの方法
元のカードにアノテーションの節が無かった。下の「調査メモの記述」と「元のカード」を参照。
## 微調整に使うときの注意
- 調査メモの台帳での区分: D(記載なし・other・unknown・資料間の食い違い)。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_kg_entity.md` | Q2. エンティティ型付け・NER・エンティティリンキング・名寄せ(同名企業の寄せ)にどんなデータがあり、Laya にどう載るか (L68) | lighteval: タグなし。wdc/products-2017: unknown。matchbench: タグなし | 確認済 | 確認印なし(各ファイルの凡例により確認済扱い) / 原文注記: Company の A 側 100 件の平均長は 7,462 文字(先頭 100 件の計測)で、先頭 100 件に正例が 0 件。小文字化・記号除去済みでノイズが多い。1 件が Laya の入力長を超えるので切り詰めが要 |
| `ontology_tables_schema_kg.md` | Q2. 2 つのスキーマ・2 つのレコードが同じものを指すか(スキーマ照合・エンティティ照合・商品の同一判定)のラベルつきデータは何があるか (L72) | タグなし | 確認済 | 原文注記: train が test より小さく、学習量は少ない。ライセンス不明 |
## 元のカード
(元のカードは空だった)