Daniel0315 commited on
Commit
869dd1c
·
verified ·
1 Parent(s): 8bda6d8

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +239 -0
README.md ADDED
@@ -0,0 +1,239 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+
3
+ # CitationHub
4
+
5
+ ## Dataset Description
6
+
7
+ CitationHub is a large-scale scholarly citation dataset designed for citation analysis, citation intent modeling, scholarly knowledge graph construction, and evidence-grounded hypothesis generation for AI co-scientist systems.
8
+
9
+ The dataset contains citation events between citing papers and highly cited seed papers, enriched with citation intent labels, citation contexts, field information, journal metadata, author and affiliation information, and heterogeneous knowledge graph nodes and edges.
10
+
11
+ The dataset is intended to support research on:
12
+
13
+ * Citation intent analysis
14
+ * Scientific impact modeling
15
+ * Future citation prediction
16
+ * Scholarly knowledge graph completion
17
+ * Intent-aware citation recommendation
18
+ * Evidence-grounded hypothesis generation
19
+ * AI co-scientist systems for scientific discovery
20
+
21
+
22
+ Dataset Repository:
23
+ https://huggingface.co/datasets/Daniel0315/CitationHub
24
+
25
+ ---
26
+
27
+ # Dataset Summary
28
+
29
+ | File | Rows | Columns | Description |
30
+ | -------------------------------------- | --------: | ------: | -------------------------------------------------------------------------------------- |
31
+ | `citation_events.parquet` | 1,857,503 | 20 | Raw citation event records |
32
+ | `citation_events_enriched.parquet` | 1,857,503 | 32 | Citation events enriched with cited seed paper metadata |
33
+ | `citation_events_normalized.parquet` | 1,857,503 | 23 | Citation events with normalized intent and field identifiers |
34
+ | `citing_papers.parquet` | 1,467,045 | 7 | Metadata of citing papers |
35
+ | `citing_papers_normalized.parquet` | 1,467,045 | 8 | Citing papers with normalized journal IDs |
36
+ | `seed_cited_papers.parquet` | 23,479 | 42 | Highly cited seed paper metadata |
37
+ | `seed_cited_papers_normalized.parquet` | 23,479 | 48 | Seed papers with normalized journal, author, affiliation, city, country, and field IDs |
38
+ | `kg_edges.parquet` | 6,855,117 | 3 | Heterogeneous knowledge graph edges |
39
+ | `kg_nodes.parquet` | 3,418,433 | 14 | Heterogeneous knowledge graph nodes |
40
+ | `authors.parquet` | 16,839 | 2 | Author ID-name mapping |
41
+ | `affiliations.parquet` | 5,271 | 2 | Affiliation ID-name mapping |
42
+ | `affiliation_geo.parquet` | 5,352 | 6 | Affiliation, city, and country mappings |
43
+ | `journals.parquet` | 46,237 | 2 | Journal ID-name mapping |
44
+ | `fields.parquet` | 21 | 3 | Field ID-name mapping |
45
+ | `intents.parquet` | 31 | 2 | Citation intent ID-name mapping |
46
+ | `cities.parquet` | 1,899 | 2 | City ID-name mapping |
47
+ | `countries.parquet` | 108 | 2 | Country ID-name mapping |
48
+
49
+ ---
50
+
51
+ # Citation Events
52
+
53
+ The main citation event files contain paper-level citation relations and citation semantics.
54
+
55
+ Important fields include:
56
+
57
+ | Field | Description |
58
+ | ----------------------- | ----------------------------------------------------------- |
59
+ | `citation_event_id` | Unique citation event identifier |
60
+ | `citing_paper_id` | Identifier of the citing paper |
61
+ | `cited_seed_paper_id` | Identifier of the cited seed paper |
62
+ | `citing_doi` | DOI of the citing paper |
63
+ | `citing_title` | Title of the citing paper |
64
+ | `citing_year` | Publication year of the citing paper |
65
+ | `citing_venue` | Venue of the citing paper |
66
+ | `primary_intent` | Primary citation intent label |
67
+ | `all_intents` | List of citation intents associated with the citation event |
68
+ | `contexts` | Citation context text snippets |
69
+ | `context_count` | Number of citation contexts |
70
+ | `intent_count` | Number of intent annotations |
71
+ | `is_influential` | Whether the citation is marked as influential |
72
+ | `has_semantic_evidence` | Whether semantic evidence is available |
73
+ | `field_id` | Normalized field identifier |
74
+
75
+ ---
76
+
77
+ # Seed Cited Papers
78
+
79
+ Seed papers represent highly cited papers used as cited targets.
80
+
81
+ Important fields include:
82
+
83
+ | Field | Description |
84
+ | ------------------- | ----------------------------------------------------------------- |
85
+ | `seed_paper_id` | Unique seed paper identifier |
86
+ | `doi` | DOI of the seed paper |
87
+ | `title` | Title of the seed paper |
88
+ | `publication_name` | Publication venue |
89
+ | `creator` | Main creator or first author |
90
+ | `citedby_count` | Citation count |
91
+ | `group` | Broad research group |
92
+ | `category` | Research category |
93
+ | `journal_clarivate` | Clarivate journal name |
94
+ | `is_seed_top5pct` | Whether the seed paper belongs to the selected highly cited group |
95
+
96
+ ---
97
+
98
+ # Knowledge Graph
99
+
100
+ The dataset includes a heterogeneous scholarly knowledge graph.
101
+
102
+ ## kg_nodes.parquet
103
+
104
+ | Field | Description |
105
+ | ------------------ | ---------------------------------------- |
106
+ | `node_id` | Unique node identifier |
107
+ | `label` | Human-readable node label |
108
+ | `node_type` | Type of node |
109
+ | `doi` | DOI if the node corresponds to a paper |
110
+ | `publication_name` | Publication venue |
111
+ | `group` | Research group |
112
+ | `citedby_count` | Citation count |
113
+ | `venue` | Venue information |
114
+ | `year` | Publication year |
115
+ | `primary_intent` | Citation intent for citation-event nodes |
116
+
117
+ ## kg_edges.parquet
118
+
119
+ | Field | Description |
120
+ | ----------- | ---------------------- |
121
+ | `source` | Source node ID |
122
+ | `target` | Target node ID |
123
+ | `edge_type` | Type of graph relation |
124
+
125
+ ---
126
+
127
+ # Intended Uses
128
+
129
+ This dataset can be used for:
130
+
131
+ 1. Citation Intent Classification
132
+ 2. Future Citation Prediction
133
+ 3. AI Co-Scientist Hypothesis Generation
134
+ 4. Scholarly Knowledge Graph Completion
135
+ 5. Scientific Opportunity Discovery
136
+ 6. Citation-based Scientific Recommendation
137
+ 7. Evidence-grounded LLM Reasoning for Scientific Discovery
138
+
139
+ ---
140
+
141
+ # Example Usage
142
+
143
+ ```python
144
+ import pandas as pd
145
+
146
+ citation_events = pd.read_parquet("citation_events_normalized.parquet")
147
+ seed_papers = pd.read_parquet("seed_cited_papers_normalized.parquet")
148
+ kg_edges = pd.read_parquet("kg_edges.parquet")
149
+ kg_nodes = pd.read_parquet("kg_nodes.parquet")
150
+
151
+ print(citation_events.shape)
152
+ print(seed_papers.shape)
153
+ print(kg_edges.shape)
154
+ print(kg_nodes.shape)
155
+ ```
156
+
157
+ ---
158
+
159
+ # Example Task: Future Citation Prediction
160
+
161
+ ```python
162
+ citation_events = pd.read_parquet("citation_events_normalized.parquet")
163
+
164
+ train = citation_events[citation_events["citing_year"] <= 2018]
165
+ test = citation_events[citation_events["citing_year"] > 2018]
166
+
167
+ print("Train:", train.shape)
168
+ print("Test:", test.shape)
169
+ ```
170
+
171
+ This task can be used as a proxy for scientific opportunity discovery: if a model can predict future citation links from past citation behavior, citation intents, and graph structure, the predicted links may indicate latent research opportunities.
172
+
173
+ ---
174
+
175
+ # Dataset Construction
176
+
177
+ The dataset was constructed by integrating:
178
+
179
+ 1. Citation event records
180
+ 2. Citation contexts and citation intent labels
181
+ 3. Highly cited seed paper metadata
182
+ 4. Citing paper metadata
183
+ 5. Journal, author, affiliation, city, country, and field normalization tables
184
+ 6. Heterogeneous scholarly knowledge graph nodes and edges
185
+
186
+ The normalized files provide ID-based mappings for relational and graph-based experiments.
187
+
188
+ ---
189
+
190
+ # Dataset Structure
191
+
192
+ ```text
193
+ CitationHub/
194
+ ├── affiliation_geo.parquet
195
+ ├── affiliations.parquet
196
+ ├── authors.parquet
197
+ ├── citation_events.parquet
198
+ ├── citation_events_enriched.parquet
199
+ ├── citation_events_normalized.parquet
200
+ ├── cities.parquet
201
+ ├── citing_papers.parquet
202
+ ├── citing_papers_normalized.parquet
203
+ ├── countries.parquet
204
+ ├── fields.parquet
205
+ ├── intents.parquet
206
+ ├── journals.parquet
207
+ ├── kg_edges.parquet
208
+ ├── kg_nodes.parquet
209
+ ├── seed_cited_papers.parquet
210
+ └── seed_cited_papers_normalized.parquet
211
+ ```
212
+
213
+
214
+
215
+ # Ethical Considerations
216
+
217
+ This dataset is intended for scholarly analysis and research support. Citation-based metrics should not be used as the sole basis for evaluating researchers, institutions, journals, or countries.
218
+
219
+ ---
220
+
221
+ # Citation
222
+
223
+ ```bibtex
224
+ @dataset{citationhub2026,
225
+ title={CitationHub: An Intent-aware Scholarly Citation Knowledge Graph Dataset},
226
+ author={Nam, Seohyun},
227
+ year={2026},
228
+ publisher={Hugging Face},
229
+ url={https://huggingface.co/datasets/Daniel0315/CitationHub}
230
+ }
231
+ ```
232
+
233
+ ---
234
+
235
+ # License
236
+
237
+ CC BY 4.0
238
+
239
+ ---