1602353775wzj commited on
Commit
9cabd9a
·
verified ·
1 Parent(s): 327c6f5

Add ClassicalPoetryAllusionsRetrieval dataset (corpus/queries/qrels)

Browse files
Files changed (1) hide show
  1. README.md +81 -45
README.md CHANGED
@@ -15,73 +15,109 @@ configs:
15
  dataset_info:
16
  configs:
17
  - config_name: corpus
 
 
18
  features:
19
- id:
20
- dtype: string
21
- sentence:
22
- dtype: string
23
- keyword:
24
- dtype: string
25
- source:
26
- dtype: string
27
  splits:
28
  - name: test
29
  num_bytes: 382933
30
  num_examples: 4838
31
- download_size: 382933
32
- dataset_size: 382933
33
  - config_name: queries
 
 
34
  features:
35
- id:
36
- dtype: string
37
- Source_Term:
38
- dtype: string
39
- Source_Content:
40
- dtype: string
41
- Derived_List:
42
- dtype: string
43
  splits:
44
  - name: test
45
  num_bytes: 1687871
46
  num_examples: 503
47
- download_size: 1687871
48
- dataset_size: 1687871
49
  - config_name: qrels
 
 
50
  features:
51
- query-id:
52
- dtype: string
53
- corpus-id:
54
- dtype: string
55
- score:
56
- dtype: float64
57
  splits:
58
  - name: test
59
  num_bytes: 36766
60
  num_examples: 5013
61
- download_size: 36766
62
- dataset_size: 36766
63
- task_categories:
64
- - text-retrieval
65
- language:
66
- - zh
67
- pretty_name: Classical Poetry Allusions Retrieval
68
  ---
69
 
70
  # Classical Poetry Allusions Retrieval Dataset
71
 
72
- ## 介绍
73
- 这是一个古汉语典故检索数据集,旨在评估模型在古代诗歌语境下对典故例句的检索能力。
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
74
 
75
- ## 数据结构
76
- 该数据集包含三个子集Configs
77
- - **corpus**: 包含所有待检索的典故例句。
78
- - **queries**: 每个典故的背景知识作为查询。
79
- - **qrels**: 查询与例句之间的相关性映射(Ground Truth)。
80
 
81
- ## 使方法
82
 
83
- python
84
- from datasets import load_dataset
85
- 加载语料库
86
- corpus = load_dataset("PoetryMTEB/ClassicalPoetryAllusionsRetrieval", "corpus", split="test")
87
 
 
 
 
 
 
 
 
 
 
 
15
  dataset_info:
16
  configs:
17
  - config_name: corpus
18
+ dataset_size: 382933
19
+ download_size: 382933
20
  features:
21
+ - name: id
22
+ dtype: string
23
+ - name: sentence
24
+ dtype: string
25
+ - name: keyword
26
+ dtype: string
27
+ - name: source
28
+ dtype: string
29
  splits:
30
  - name: test
31
  num_bytes: 382933
32
  num_examples: 4838
 
 
33
  - config_name: queries
34
+ dataset_size: 1687871
35
+ download_size: 1687871
36
  features:
37
+ - name: id
38
+ dtype: string
39
+ - name: Source_Term
40
+ dtype: string
41
+ - name: Source_Content
42
+ dtype: string
43
+ - name: Derived_List
44
+ dtype: string
45
  splits:
46
  - name: test
47
  num_bytes: 1687871
48
  num_examples: 503
 
 
49
  - config_name: qrels
50
+ dataset_size: 36766
51
+ download_size: 36766
52
  features:
53
+ - name: query-id
54
+ dtype: string
55
+ - name: corpus-id
56
+ dtype: string
57
+ - name: score
58
+ dtype: float64
59
  splits:
60
  - name: test
61
  num_bytes: 36766
62
  num_examples: 5013
 
 
 
 
 
 
 
63
  ---
64
 
65
  # Classical Poetry Allusions Retrieval Dataset
66
 
67
+ ## 数据集描述
68
+
69
+ 这是一个**古汉语典故检索(Retrieval)评测数据集**,包含三部分:
70
+
71
+ - **corpus**:典故例句语料库(每条为一个诗歌/文本中的例句)
72
+ - **queries**:每个典故对应一个查询(来自典故知识库的“典源词/典源内容/衍生列表”)
73
+ - **qrels**:相关性标注(query 与其对应的例句都标注为相关,`score=1`)
74
+
75
+ ## 数据来源与构建方法
76
+
77
+ - **数据源**:典故知识库与标注语料来自开源项目 [Ancient-Chinese-Allusion-Resource-Database](https://github.com/KaijieMo-kj/Ancient-Chinese-Allusion-Resource-Database)。
78
+ - **采样方法**:对每个典故优先采样 `keyword` 尽可能不同的例子,若某典故例子数 < 10 则不采样;每个保留典故最多采样 10 条例句。
79
+ - **相关性定义**:同一典故的查询与其采样到的例句均为相关(`score=1`)。
80
+
81
+ ## 数据集统计(test split)
82
+
83
+ | 子集 | 样本数量 | 文件大小 | 描述 |
84
+ |------|----------|----------|------|
85
+ | **corpus** | 4838 | 382933 bytes | 例句文档库 |
86
+ | **queries** | 503 | 1687871 bytes | 典故查询集合 |
87
+ | **qrels** | 5013 | 36766 bytes | 相关性标注 |
88
+
89
+ ## 数据字段说明
90
+
91
+ ### corpus
92
+
93
+ - `id`: 例句唯一标识符(来自原始例句 id)
94
+ - `sentence`: 例句文本
95
+ - `keyword`: 例句中触发该典故的关键词(采样时用于尽量多样化)
96
+ - `source`: 例句来源
97
+
98
+ ### queries
99
+
100
+ - `id`: 查询唯一标识符(顺序编码,字符串)
101
+ - `Source_Term`: 典源词(典故名)
102
+ - `Source_Content`: 典源内容(来源与出处)
103
+ - `Derived_List`: 衍生列表(JSON 字符串,保留完整结构)
104
+
105
+ ### qrels
106
 
107
+ - `query-id`: 查询 id(与 `queries.id` 对应)
108
+ - `corpus-id`: 文档 id与 `corpus.id` 对应
109
+ - `score`: 相关性分数(全部为 1)
 
 
110
 
111
+ ## 信息
112
 
113
+ 如果你在研究中使用了该数据源,请引用原项目提供的论文:
 
 
 
114
 
115
+ ```bibtex
116
+ @article{Kaijie2024,
117
+ title={古汉语典故资源库的构建及应用研究},
118
+ author={莫凯洁,丘子靓,王予沛,胡韧奋},
119
+ journal={中文信息学报},
120
+ year={2024},
121
+ pages = {27--34}
122
+ }
123
+ ```