HYUKI5 commited on
Commit
c97dfa6
·
verified ·
1 Parent(s): fc2b0ef

tasksource/comve@3bf449686b1f を複製し、カードを日本語で書く

Browse files
README.md ADDED
@@ -0,0 +1,140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-sa-4.0
3
+ pretty_name: tasksource/comve (mirror)
4
+ source_datasets:
5
+ - tasksource/comve
6
+ tags:
7
+ - laya
8
+ - laya-finetune
9
+ - mirror
10
+ configs:
11
+ - config_name: default
12
+ data_files:
13
+ - split: test
14
+ path: data/test-*
15
+ - split: validation
16
+ path: data/validation-*
17
+ - split: train
18
+ path: data/train-*
19
+ dataset_info:
20
+ features:
21
+ - name: inputs
22
+ dtype: string
23
+ - name: choices_list
24
+ list: string
25
+ - name: labels
26
+ dtype: int64
27
+ - name: metadata
28
+ dtype: string
29
+ splits:
30
+ - name: test
31
+ num_bytes: 349165
32
+ num_examples: 999
33
+ - name: validation
34
+ num_bytes: 358552
35
+ num_examples: 997
36
+ - name: train
37
+ num_bytes: 3553117
38
+ num_examples: 9966
39
+ download_size: 1925400
40
+ dataset_size: 4260834
41
+ ---
42
+
43
+ # tasksource/comve(複製)
44
+
45
+ ## データの内容
46
+
47
+ SemEval-2020 Task 4 の ComVE サブタスク A の、常識に合う文と合わない文を判定するためのデータ。1 行が文(`inputs`)と選択肢の列(`choices_list`)、正解の番号(`labels`)、付随情報(`metadata`)からなる。公式の CSV を結合し、無意味のラベルを反転させ、元からある holdout を取り分けて作った。生成した選択肢や負例は含まず、元の README は CC BY-SA 4.0 と明記している。
48
+
49
+ ## データの概要
50
+
51
+ | 項目 | 値 |
52
+ |---|---|
53
+ | 行数 | 11,962(1 構成・3 分割の合計) |
54
+ | 列数 | 4 |
55
+ | 容量 | 1.93 MB(ファイル 7 本) |
56
+ | ファイルの形式 | parquet 3、jsonl 1、json 1 |
57
+ | 公開 | public |
58
+
59
+ ### 構成と分割
60
+
61
+ | 構成(config) | 分割(split)ごとの行数 | 列数 |
62
+ |---|---|---:|
63
+ | `default` | `test` 999、`validation` 997、`train` 9,966 | 4 |
64
+
65
+ ### 列(構成 `default`・分割 `train`)
66
+
67
+ | 列 | 型 | 先頭の行の値 |
68
+ |---|---|---|
69
+ | `inputs` | `string` | `Which statement is more sensible in ordinary cir…` |
70
+ | `choices_list` | `list<string>` | `[2 件]` |
71
+ | `labels` | `int64` | `1` |
72
+ | `metadata` | `string` | `{"annotation_origin": "human", "id": "train:0", …` |
73
+
74
+ 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。
75
+
76
+ ## 複製について
77
+
78
+ [`tasksource/comve`](https://huggingface.co/datasets/tasksource/comve) をリビジョン `3bf449686b1f3c2867aafecb2237ae5c9027e240` 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。
79
+
80
+ ## 出典
81
+
82
+ - 元のデータセット: https://huggingface.co/datasets/tasksource/comve
83
+ - 複製したリビジョン: `3bf449686b1f3c2867aafecb2237ae5c9027e240`(2026-10-15 09:47:54+00:00)
84
+ - 元のライセンス表記: `cc-by-sa-4.0`
85
+
86
+ ## 論文
87
+
88
+ - 元のカードに論文へのリンクは見つからなかった
89
+
90
+ ## アノテーションの方法
91
+
92
+ (元のカードに該当の見出しが無いので、ラベルの付け方に触れた段落を抜き出した)
93
+
94
+ 元の配布先: https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation
95
+ 元の README はライセンスを CC BY-SA 4.0 と明記している。生成した選択肢や負例は含まない。
96
+
97
+ ## 微調整に使うときの注意
98
+
99
+ - 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。
100
+ - 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。
101
+ - 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。
102
+
103
+ ## 調査メモの記述
104
+
105
+ | メモ | 節 | ライセンス(記載) | 確認 | 注記 |
106
+ |---|---|---|---|---|
107
+ | `jev_sources_tasksource_bekko.csv` | CSV(tasksource-jev-typed-decisions) | cc-by-sa-4.0, cc-by-sa-4.0 (original source) / license_use=commercial | 確認済 | license_use=commercial / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
108
+
109
+ ## 元のカード(日本語訳)
110
+
111
+ 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。
112
+
113
+ ---
114
+
115
+ # comve
116
+
117
+ 公式の ComVE サブタスク A の CSV を結合し、無意味のラベルを反転させ、元からある holdout を取り分けておく。
118
+
119
+ 元の配布先: https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation
120
+ 元の README はライセンスを CC BY-SA 4.0 と明記している。生成した選択肢や負例は含まない。
121
+
122
+ 元のデータ: [元の配布先](https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation)。[tasksource](https://github.com/sileod/tasksource) 向けに [scripts/repackage_dataset/comve.py](https://github.com/sileod/tasksource/blob/main/scripts/repackage_dataset/comve.py#L56) で parquet に再梱包した。
123
+
124
+ [変換の来歴](provenance.json): 元の版、ハッシュ、設定、除外の内訳。
125
+
126
+ <details>
127
+ <summary>原文(英語)</summary>
128
+
129
+ # comve
130
+
131
+ Join official ComVE subtask-A CSVs, invert nonsense labels, reserve native holdouts.
132
+
133
+ Original distribution: https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation
134
+ Original README declares CC BY-SA 4.0. No generated choices or negatives.
135
+
136
+ Original data: [original distribution](https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/repackage_dataset/comve.py](https://github.com/sileod/tasksource/blob/main/scripts/repackage_dataset/comve.py#L56).
137
+
138
+ [Conversion provenance](provenance.json): source revisions, hashes, configuration and exclusions.
139
+
140
+ </details>
data/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e1d360a1a4a7fcbd56cf82e581f99f2f92d916d2b0908b7b42cdb9192cc2b75
3
+ size 160896
data/train-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1b5493eea8409c19a7622d25968f90841c3bd6a29217aa5a4d037edde2a6b83
3
+ size 1605429
data/validation-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd82af5c7a2616acffef67012354c70f148ec76d8bf5654d22ae731fdfee0780
3
+ size 159075
excluded-questions.jsonl ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"split": "test", "id": "635", "reason": "duplicate_pair"}
2
+ {"split": "train", "id": "553", "reason": "duplicate_pair"}
3
+ {"split": "train", "id": "2077", "reason": "duplicate_pair"}
4
+ {"split": "train", "id": "2251", "reason": "duplicate_pair"}
5
+ {"split": "train", "id": "2559", "reason": "duplicate_pair"}
6
+ {"split": "train", "id": "3543", "reason": "duplicate_pair"}
7
+ {"split": "train", "id": "3714", "reason": "duplicate_pair"}
8
+ {"split": "train", "id": "4075", "reason": "invalid_choices"}
9
+ {"split": "train", "id": "4121", "reason": "invalid_choices"}
10
+ {"split": "train", "id": "4279", "reason": "duplicate_pair"}
11
+ {"split": "train", "id": "4313", "reason": "invalid_choices"}
12
+ {"split": "train", "id": "4965", "reason": "conflicting_gold"}
13
+ {"split": "train", "id": "5697", "reason": "duplicate_pair"}
14
+ {"split": "train", "id": "5698", "reason": "duplicate_pair"}
15
+ {"split": "train", "id": "5699", "reason": "duplicate_pair"}
16
+ {"split": "train", "id": "5700", "reason": "duplicate_pair"}
17
+ {"split": "train", "id": "5703", "reason": "duplicate_pair"}
18
+ {"split": "train", "id": "5795", "reason": "duplicate_pair"}
19
+ {"split": "train", "id": "6398", "reason": "invalid_choices"}
20
+ {"split": "train", "id": "6954", "reason": "duplicate_pair"}
21
+ {"split": "train", "id": "7047", "reason": "duplicate_pair"}
22
+ {"split": "train", "id": "7700", "reason": "invalid_choices"}
23
+ {"split": "train", "id": "7931", "reason": "duplicate_pair"}
24
+ {"split": "train", "id": "7957", "reason": "conflicting_gold"}
25
+ {"split": "train", "id": "8130", "reason": "duplicate_pair"}
26
+ {"split": "train", "id": "8420", "reason": "duplicate_pair"}
27
+ {"split": "train", "id": "8760", "reason": "duplicate_pair"}
28
+ {"split": "train", "id": "8773", "reason": "duplicate_pair"}
29
+ {"split": "train", "id": "8774", "reason": "duplicate_pair"}
30
+ {"split": "train", "id": "8776", "reason": "duplicate_pair"}
31
+ {"split": "train", "id": "8779", "reason": "duplicate_pair"}
32
+ {"split": "train", "id": "8780", "reason": "duplicate_pair"}
33
+ {"split": "train", "id": "8786", "reason": "duplicate_pair"}
34
+ {"split": "train", "id": "9459", "reason": "duplicate_pair"}
35
+ {"split": "train", "id": "9965", "reason": "duplicate_pair"}
provenance.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "source": "https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation",
3
+ "revision": "c55b1a9e6cb66997675f85073e17a9522cf38c2c",
4
+ "source_sha256": {
5
+ "ALL data/Training Data/subtaskA_data_all.csv": "e0f0dca581e92d6f97a819f44a3157058741e1c9dc92fd036f91bf3adf382215",
6
+ "ALL data/Training Data/subtaskA_answers_all.csv": "8a30a09dcfcdc536d03201bea976708dff3813691667a3c1a346af15bfabbc69",
7
+ "ALL data/Dev Data/subtaskA_dev_data.csv": "60f32177d667016e4ca2513fe75113f4e30b8336e3d3c5b8de4bcedfa53571bc",
8
+ "ALL data/Dev Data/subtaskA_gold_answers.csv": "405e8a28b1d79c19deb7855d44c46e9ac2e5d0d09dd2280476b21341fa492a95",
9
+ "ALL data/Test Data/subtaskA_test_data.csv": "801e49314f3d6b2dcb72f30cce58d1212b4d84c1fc9b7f1e1da27c6285a32814",
10
+ "ALL data/Test Data/subtaskA_gold_answers.csv": "a0fb5a54d788975168d87e058339aa490842e89c94d0719d2ca3016b47a39836"
11
+ },
12
+ "license": "cc-by-sa-4.0",
13
+ "annotation_origin": "human",
14
+ "native_counts": {
15
+ "train": 10000,
16
+ "validation": 997,
17
+ "test": 1000
18
+ },
19
+ "excluded_counts": {
20
+ "test": {
21
+ "duplicate_pair": 1
22
+ },
23
+ "validation": {},
24
+ "train": {
25
+ "duplicate_pair": 27,
26
+ "invalid_choices": 5,
27
+ "conflicting_gold": 2
28
+ }
29
+ },
30
+ "conversion_sha256": "ccc0a2f4fc29716c1bb140b511cfcf7210ab72e8958d6718635747e591a52f13",
31
+ "configuration": {
32
+ "gold": "invert native nonsensical index",
33
+ "split_policy": "native; test > validation > train"
34
+ },
35
+ "splits": {
36
+ "test": 999,
37
+ "validation": 997,
38
+ "train": 9966
39
+ },
40
+ "label_counts": {
41
+ "test": {
42
+ "1": 508,
43
+ "0": 491
44
+ },
45
+ "validation": {
46
+ "1": 518,
47
+ "0": 479
48
+ },
49
+ "train": {
50
+ "1": 4961,
51
+ "0": 5005
52
+ }
53
+ }
54
+ }