Commit
·
a245956
1
Parent(s):
c434d86
Track important_citations.csv with Git LFS
Browse files- .gitattributes +1 -0
- README.md +5 -4
- important_citations.csv +3 -0
- lotus_deep_research.py +397 -0
- papers.csv +0 -0
- papers_with_related_works.csv +0 -0
- usage_example.py +175 -0
.gitattributes
CHANGED
@@ -57,3 +57,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
57 |
# Video files - compressed
|
58 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
59 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
|
|
|
57 |
# Video files - compressed
|
58 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
59 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
60 |
+
important_citations.csv filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
@@ -23,7 +23,7 @@ A comprehensive dataset of academic papers with extracted related works sections
|
|
23 |
|
24 |
## 📊 Dataset Overview
|
25 |
|
26 |
-
This dataset contains **
|
27 |
|
28 |
### 🎯 Use Cases
|
29 |
|
@@ -35,10 +35,9 @@ This dataset contains **67 academic papers** from ArXiv with their related works
|
|
35 |
|
36 |
## 📁 Dataset Structure
|
37 |
|
38 |
-
### 1. `papers_with_related_works.csv` (
|
39 |
|
40 |
Contains academic papers with extracted related works sections in multiple formats:
|
41 |
-
|
42 |
| Column | Description |
|
43 |
|--------|-------------|
|
44 |
| `arxiv_id` | ArXiv identifier (e.g., "2506.02838v1") |
|
@@ -50,11 +49,12 @@ Contains academic papers with extracted related works sections in multiple forma
|
|
50 |
| `updated_date` | Last update date |
|
51 |
| `abs_url` | ArXiv abstract URL |
|
52 |
| `arxiv_link` | Full ArXiv link |
|
|
|
53 |
| `raw_latex_related_works` | Raw LaTeX related works section |
|
54 |
| `clean_latex_related_works` | Cleaned LaTeX related works section |
|
55 |
| `pdf_related_works` | Related works extracted from PDF |
|
56 |
|
57 |
-
### 2. `
|
58 |
|
59 |
Contains individual citations with recovered metadata:
|
60 |
|
@@ -75,6 +75,7 @@ Contains individual citations with recovered metadata:
|
|
75 |
| `bib_paper_url` | URL of the cited paper |
|
76 |
| `bib_paper_doi` | DOI of the cited paper |
|
77 |
| `bib_paper_journal` | Journal name |
|
|
|
78 |
| `search_res_title` | Title from search results |
|
79 |
| `search_res_url` | URL from search results |
|
80 |
| `search_res_content` | Content snippet from search results |
|
|
|
23 |
|
24 |
## 📊 Dataset Overview
|
25 |
|
26 |
+
This dataset contains **63 academic papers** from ArXiv with their related works sections and **1630 recovered citations**, providing a rich resource for research generation and citation analysis tasks.
|
27 |
|
28 |
### 🎯 Use Cases
|
29 |
|
|
|
35 |
|
36 |
## 📁 Dataset Structure
|
37 |
|
38 |
+
### 1. `papers_with_related_works.csv` (63 papers)
|
39 |
|
40 |
Contains academic papers with extracted related works sections in multiple formats:
|
|
|
41 |
| Column | Description |
|
42 |
|--------|-------------|
|
43 |
| `arxiv_id` | ArXiv identifier (e.g., "2506.02838v1") |
|
|
|
49 |
| `updated_date` | Last update date |
|
50 |
| `abs_url` | ArXiv abstract URL |
|
51 |
| `arxiv_link` | Full ArXiv link |
|
52 |
+
| `publication_date` | Publication date |
|
53 |
| `raw_latex_related_works` | Raw LaTeX related works section |
|
54 |
| `clean_latex_related_works` | Cleaned LaTeX related works section |
|
55 |
| `pdf_related_works` | Related works extracted from PDF |
|
56 |
|
57 |
+
### 2. `recovered_citations.csv` (1630 citations)
|
58 |
|
59 |
Contains individual citations with recovered metadata:
|
60 |
|
|
|
75 |
| `bib_paper_url` | URL of the cited paper |
|
76 |
| `bib_paper_doi` | DOI of the cited paper |
|
77 |
| `bib_paper_journal` | Journal name |
|
78 |
+
| `original_title` | Original title from citation metadata |
|
79 |
| `search_res_title` | Title from search results |
|
80 |
| `search_res_url` | URL from search results |
|
81 |
| `search_res_content` | Content snippet from search results |
|
important_citations.csv
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5cb48008cdaa202abc4ed2e6689357f1858791b5fe0d586b33f9b447797114cf
|
3 |
+
size 16641640
|
lotus_deep_research.py
ADDED
@@ -0,0 +1,397 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
Lotus Deep Research Dataset: Academic papers with extracted related works sections and recovered citations.
|
3 |
+
|
4 |
+
This dataset contains academic papers from ArXiv with their related works sections and recovered citations,
|
5 |
+
providing a rich resource for research generation and citation analysis tasks.
|
6 |
+
"""
|
7 |
+
|
8 |
+
import csv
|
9 |
+
import datasets
|
10 |
+
from typing import Dict, List, Any, Optional
|
11 |
+
|
12 |
+
|
13 |
+
# Dataset URLs - these would typically point to hosted files
|
14 |
+
_DESCRIPTION = """\
|
15 |
+
A comprehensive dataset of academic papers with extracted related works sections and recovered citations,
|
16 |
+
designed for training and evaluating research generation systems.
|
17 |
+
|
18 |
+
This dataset contains 63 academic papers from ArXiv with their related works sections and 1630 recovered citations,
|
19 |
+
providing a rich resource for research generation and citation analysis tasks.
|
20 |
+
"""
|
21 |
+
|
22 |
+
_CITATION = """\
|
23 |
+
@misc{patel2025deepscholarbenchlivebenchmarkautomated,
|
24 |
+
title={DeepScholar-Bench: A Live Benchmark and Automated Evaluation for Generative Research Synthesis},
|
25 |
+
author={Liana Patel and Negar Arabzadeh and Harshit Gupta and Ankita Sundar and Ion Stoica and Matei Zaharia and Carlos Guestrin},
|
26 |
+
year={2025},
|
27 |
+
eprint={2508.20033},
|
28 |
+
archivePrefix={arXiv},
|
29 |
+
primaryClass={cs.CL},
|
30 |
+
url={https://arxiv.org/abs/2508.20033},
|
31 |
+
}
|
32 |
+
"""
|
33 |
+
|
34 |
+
_HOMEPAGE = "https://github.com/guestrin-lab/deepscholar-bench"
|
35 |
+
_LICENSE = "MIT"
|
36 |
+
|
37 |
+
# URLs to the dataset files
|
38 |
+
_URLS = {
|
39 |
+
"papers": "papers_with_related_works.csv",
|
40 |
+
"citations": "recovered_citations.csv",
|
41 |
+
"important_citations": "important_citations.csv",
|
42 |
+
}
|
43 |
+
|
44 |
+
|
45 |
+
class LotusDeepResearchConfig(datasets.BuilderConfig):
|
46 |
+
"""BuilderConfig for LotusDeepResearch dataset."""
|
47 |
+
|
48 |
+
def __init__(self, name: str, description: str, **kwargs):
|
49 |
+
"""BuilderConfig for LotusDeepResearch.
|
50 |
+
|
51 |
+
Args:
|
52 |
+
name: Configuration name
|
53 |
+
description: Description of this configuration
|
54 |
+
**kwargs: Additional keyword arguments
|
55 |
+
"""
|
56 |
+
super(LotusDeepResearchConfig, self).__init__(
|
57 |
+
name=name,
|
58 |
+
description=description,
|
59 |
+
version=datasets.Version("1.0.0"),
|
60 |
+
**kwargs
|
61 |
+
)
|
62 |
+
|
63 |
+
|
64 |
+
class LotusDeepResearch(datasets.GeneratorBasedBuilder):
|
65 |
+
"""Lotus Deep Research dataset."""
|
66 |
+
|
67 |
+
VERSION = datasets.Version("1.0.0")
|
68 |
+
|
69 |
+
BUILDER_CONFIGS = [
|
70 |
+
LotusDeepResearchConfig(
|
71 |
+
name="papers",
|
72 |
+
description="Academic papers with extracted related works sections (63 papers)",
|
73 |
+
),
|
74 |
+
LotusDeepResearchConfig(
|
75 |
+
name="citations",
|
76 |
+
description="Recovered citations with metadata (1630 citations)",
|
77 |
+
),
|
78 |
+
LotusDeepResearchConfig(
|
79 |
+
name="important_citations",
|
80 |
+
description="Important citations with metadata (1050 citations)",
|
81 |
+
),
|
82 |
+
LotusDeepResearchConfig(
|
83 |
+
name="full",
|
84 |
+
description="Complete dataset with both papers and citations",
|
85 |
+
),
|
86 |
+
]
|
87 |
+
|
88 |
+
DEFAULT_CONFIG_NAME = "full"
|
89 |
+
|
90 |
+
def _info(self) -> datasets.DatasetInfo:
|
91 |
+
"""Return the dataset info."""
|
92 |
+
|
93 |
+
if self.config.name == "papers":
|
94 |
+
features = datasets.Features({
|
95 |
+
"arxiv_id": datasets.Value("string"),
|
96 |
+
"title": datasets.Value("string"),
|
97 |
+
"authors": datasets.Value("string"),
|
98 |
+
"abstract": datasets.Value("string"),
|
99 |
+
"categories": datasets.Value("string"),
|
100 |
+
"published_date": datasets.Value("string"),
|
101 |
+
"updated_date": datasets.Value("string"),
|
102 |
+
"abs_url": datasets.Value("string"),
|
103 |
+
"arxiv_link": datasets.Value("string"),
|
104 |
+
"publication_date": datasets.Value("string"),
|
105 |
+
"raw_latex_related_works": datasets.Value("string"),
|
106 |
+
"clean_latex_related_works": datasets.Value("string"),
|
107 |
+
"pdf_related_works": datasets.Value("string"),
|
108 |
+
})
|
109 |
+
elif self.config.name == "citations":
|
110 |
+
features = datasets.Features({
|
111 |
+
"parent_paper_title": datasets.Value("string"),
|
112 |
+
"parent_paper_arxiv_id": datasets.Value("string"),
|
113 |
+
"citation_shorthand": datasets.Value("string"),
|
114 |
+
"raw_citation_text": datasets.Value("string"),
|
115 |
+
"cited_paper_title": datasets.Value("string"),
|
116 |
+
"cited_paper_arxiv_link": datasets.Value("string"),
|
117 |
+
"cited_paper_abstract": datasets.Value("string"),
|
118 |
+
"has_metadata": datasets.Value("bool"),
|
119 |
+
"is_arxiv_paper": datasets.Value("bool"),
|
120 |
+
"bib_paper_authors": datasets.Value("string"),
|
121 |
+
"bib_paper_year": datasets.Value("float32"),
|
122 |
+
"bib_paper_month": datasets.Value("string"),
|
123 |
+
"bib_paper_url": datasets.Value("string"),
|
124 |
+
"bib_paper_doi": datasets.Value("string"),
|
125 |
+
"bib_paper_journal": datasets.Value("string"),
|
126 |
+
"original_title": datasets.Value("string"),
|
127 |
+
"search_res_title": datasets.Value("string"),
|
128 |
+
"search_res_url": datasets.Value("string"),
|
129 |
+
"search_res_content": datasets.Value("string"),
|
130 |
+
})
|
131 |
+
elif self.config.name == "important_citations":
|
132 |
+
features = datasets.Features({
|
133 |
+
"parent_paper_title": datasets.Value("string"),
|
134 |
+
"parent_paper_arxiv_id": datasets.Value("string"),
|
135 |
+
"citation_shorthand": datasets.Value("string"),
|
136 |
+
"raw_citation_text": datasets.Value("string"),
|
137 |
+
"cited_paper_title": datasets.Value("string"),
|
138 |
+
"cited_paper_arxiv_link": datasets.Value("string"),
|
139 |
+
"cited_paper_abstract": datasets.Value("string"),
|
140 |
+
"has_metadata": datasets.Value("bool"),
|
141 |
+
"is_arxiv_paper": datasets.Value("bool"),
|
142 |
+
"cited_paper_authors": datasets.Value("string"),
|
143 |
+
"bib_paper_year": datasets.Value("float32"),
|
144 |
+
"bib_paper_month": datasets.Value("string"),
|
145 |
+
"bib_paper_url": datasets.Value("string"),
|
146 |
+
"bib_paper_doi": datasets.Value("string"),
|
147 |
+
"bib_paper_journal": datasets.Value("string"),
|
148 |
+
"original_title": datasets.Value("string"),
|
149 |
+
"search_res_title": datasets.Value("string"),
|
150 |
+
"search_res_url": datasets.Value("string"),
|
151 |
+
"search_res_content": datasets.Value("string"),
|
152 |
+
"arxiv_id": datasets.Value("string"),
|
153 |
+
"arxiv_link": datasets.Value("string"),
|
154 |
+
"publication_date": datasets.Value("string"),
|
155 |
+
"title": datasets.Value("string"),
|
156 |
+
"abstract": datasets.Value("string"),
|
157 |
+
"raw_latex_related_works": datasets.Value("string"),
|
158 |
+
"related_work_section": datasets.Value("string"),
|
159 |
+
"pdf_related_works": datasets.Value("string"),
|
160 |
+
"cited_paper_content": datasets.Value("string"),
|
161 |
+
})
|
162 |
+
else: # full config
|
163 |
+
features = datasets.Features({
|
164 |
+
# Papers features
|
165 |
+
"papers": datasets.Sequence({
|
166 |
+
"arxiv_id": datasets.Value("string"),
|
167 |
+
"title": datasets.Value("string"),
|
168 |
+
"authors": datasets.Value("string"),
|
169 |
+
"abstract": datasets.Value("string"),
|
170 |
+
"categories": datasets.Value("string"),
|
171 |
+
"published_date": datasets.Value("string"),
|
172 |
+
"updated_date": datasets.Value("string"),
|
173 |
+
"abs_url": datasets.Value("string"),
|
174 |
+
"arxiv_link": datasets.Value("string"),
|
175 |
+
"publication_date": datasets.Value("string"),
|
176 |
+
"raw_latex_related_works": datasets.Value("string"),
|
177 |
+
"clean_latex_related_works": datasets.Value("string"),
|
178 |
+
"pdf_related_works": datasets.Value("string"),
|
179 |
+
}),
|
180 |
+
# Citations features
|
181 |
+
"citations": datasets.Sequence({
|
182 |
+
"parent_paper_title": datasets.Value("string"),
|
183 |
+
"parent_paper_arxiv_id": datasets.Value("string"),
|
184 |
+
"citation_shorthand": datasets.Value("string"),
|
185 |
+
"raw_citation_text": datasets.Value("string"),
|
186 |
+
"cited_paper_title": datasets.Value("string"),
|
187 |
+
"cited_paper_arxiv_link": datasets.Value("string"),
|
188 |
+
"cited_paper_abstract": datasets.Value("string"),
|
189 |
+
"has_metadata": datasets.Value("bool"),
|
190 |
+
"is_arxiv_paper": datasets.Value("bool"),
|
191 |
+
"bib_paper_authors": datasets.Value("string"),
|
192 |
+
"bib_paper_year": datasets.Value("float32"),
|
193 |
+
"bib_paper_month": datasets.Value("string"),
|
194 |
+
"bib_paper_url": datasets.Value("string"),
|
195 |
+
"bib_paper_doi": datasets.Value("string"),
|
196 |
+
"bib_paper_journal": datasets.Value("string"),
|
197 |
+
"original_title": datasets.Value("string"),
|
198 |
+
"search_res_title": datasets.Value("string"),
|
199 |
+
"search_res_url": datasets.Value("string"),
|
200 |
+
"search_res_content": datasets.Value("string"),
|
201 |
+
}),
|
202 |
+
"important_citations": datasets.Sequence({
|
203 |
+
"parent_paper_title": datasets.Value("string"),
|
204 |
+
"parent_paper_arxiv_id": datasets.Value("string"),
|
205 |
+
"citation_shorthand": datasets.Value("string"),
|
206 |
+
"raw_citation_text": datasets.Value("string"),
|
207 |
+
"cited_paper_title": datasets.Value("string"),
|
208 |
+
"cited_paper_arxiv_link": datasets.Value("string"),
|
209 |
+
"cited_paper_abstract": datasets.Value("string"),
|
210 |
+
"has_metadata": datasets.Value("bool"),
|
211 |
+
"is_arxiv_paper": datasets.Value("bool"),
|
212 |
+
"cited_paper_authors": datasets.Value("string"),
|
213 |
+
"bib_paper_year": datasets.Value("float32"),
|
214 |
+
"bib_paper_month": datasets.Value("string"),
|
215 |
+
"bib_paper_url": datasets.Value("string"),
|
216 |
+
"bib_paper_doi": datasets.Value("string"),
|
217 |
+
"bib_paper_journal": datasets.Value("string"),
|
218 |
+
"original_title": datasets.Value("string"),
|
219 |
+
"search_res_title": datasets.Value("string"),
|
220 |
+
"search_res_url": datasets.Value("string"),
|
221 |
+
"search_res_content": datasets.Value("string"),
|
222 |
+
"arxiv_id": datasets.Value("string"),
|
223 |
+
"arxiv_link": datasets.Value("string"),
|
224 |
+
"publication_date": datasets.Value("string"),
|
225 |
+
"title": datasets.Value("string"),
|
226 |
+
"abstract": datasets.Value("string"),
|
227 |
+
"raw_latex_related_works": datasets.Value("string"),
|
228 |
+
"related_work_section": datasets.Value("string"),
|
229 |
+
"pdf_related_works": datasets.Value("string"),
|
230 |
+
"cited_paper_content": datasets.Value("string"),
|
231 |
+
})
|
232 |
+
})
|
233 |
+
|
234 |
+
return datasets.DatasetInfo(
|
235 |
+
description=_DESCRIPTION,
|
236 |
+
features=features,
|
237 |
+
homepage=_HOMEPAGE,
|
238 |
+
license=_LICENSE,
|
239 |
+
citation=_CITATION,
|
240 |
+
)
|
241 |
+
|
242 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
243 |
+
"""Return the dataset splits."""
|
244 |
+
|
245 |
+
# For local files, use the actual file paths
|
246 |
+
import os
|
247 |
+
current_dir = os.path.dirname(os.path.abspath(__file__))
|
248 |
+
|
249 |
+
if self.config.name == "papers":
|
250 |
+
data_file = os.path.join(current_dir, "papers_with_related_works.csv")
|
251 |
+
return [
|
252 |
+
datasets.SplitGenerator(
|
253 |
+
name=datasets.Split.TRAIN,
|
254 |
+
gen_kwargs={
|
255 |
+
"filepath": data_file,
|
256 |
+
"split": "papers"
|
257 |
+
},
|
258 |
+
),
|
259 |
+
]
|
260 |
+
elif self.config.name == "citations":
|
261 |
+
data_file = os.path.join(current_dir, "recovered_citations.csv")
|
262 |
+
return [
|
263 |
+
datasets.SplitGenerator(
|
264 |
+
name=datasets.Split.TRAIN,
|
265 |
+
gen_kwargs={
|
266 |
+
"filepath": data_file,
|
267 |
+
"split": "citations"
|
268 |
+
},
|
269 |
+
),
|
270 |
+
]
|
271 |
+
elif self.config.name == "important_citations":
|
272 |
+
data_file = os.path.join(current_dir, "important_citations.csv")
|
273 |
+
return [
|
274 |
+
datasets.SplitGenerator(
|
275 |
+
name=datasets.Split.TRAIN,
|
276 |
+
gen_kwargs={
|
277 |
+
"filepath": data_file,
|
278 |
+
"split": "important_citations"
|
279 |
+
},
|
280 |
+
),
|
281 |
+
]
|
282 |
+
else: # full config
|
283 |
+
papers_file = os.path.join(current_dir, "papers_with_related_works.csv")
|
284 |
+
citations_file = os.path.join(current_dir, "recovered_citations.csv")
|
285 |
+
important_citations_file = os.path.join(current_dir, "important_citations.csv")
|
286 |
+
return [
|
287 |
+
datasets.SplitGenerator(
|
288 |
+
name="papers",
|
289 |
+
gen_kwargs={
|
290 |
+
"filepath": papers_file,
|
291 |
+
"split": "papers"
|
292 |
+
},
|
293 |
+
),
|
294 |
+
datasets.SplitGenerator(
|
295 |
+
name="citations",
|
296 |
+
gen_kwargs={
|
297 |
+
"filepath": citations_file,
|
298 |
+
"split": "citations"
|
299 |
+
},
|
300 |
+
),
|
301 |
+
datasets.SplitGenerator(
|
302 |
+
name="important_citations",
|
303 |
+
gen_kwargs={
|
304 |
+
"filepath": important_citations_file,
|
305 |
+
"split": "important_citations"
|
306 |
+
},
|
307 |
+
),
|
308 |
+
]
|
309 |
+
|
310 |
+
def _generate_examples(self, filepath: str, split: str):
|
311 |
+
"""Generate examples from the dataset."""
|
312 |
+
|
313 |
+
def _safe_bool_convert(value: str) -> bool:
|
314 |
+
"""Safely convert string to boolean."""
|
315 |
+
if isinstance(value, str):
|
316 |
+
return value.lower() in ('true', 'yes', '1')
|
317 |
+
return bool(value)
|
318 |
+
|
319 |
+
def _safe_float_convert(value: str) -> Optional[float]:
|
320 |
+
"""Safely convert string to float."""
|
321 |
+
if not value or value.strip() == '' or value.lower() == 'nan':
|
322 |
+
return None
|
323 |
+
try:
|
324 |
+
return float(value)
|
325 |
+
except (ValueError, TypeError):
|
326 |
+
return None
|
327 |
+
|
328 |
+
if split == "papers":
|
329 |
+
with open(filepath, encoding="utf-8") as f:
|
330 |
+
reader = csv.DictReader(f)
|
331 |
+
for key, row in enumerate(reader):
|
332 |
+
yield key, {
|
333 |
+
"arxiv_id": row.get("arxiv_id", ""),
|
334 |
+
"title": row.get("title", ""),
|
335 |
+
"authors": row.get("authors", ""),
|
336 |
+
"abstract": row.get("abstract", ""),
|
337 |
+
"categories": row.get("categories", ""),
|
338 |
+
"published_date": row.get("published_date", ""),
|
339 |
+
"updated_date": row.get("updated_date", ""),
|
340 |
+
"abs_url": row.get("abs_url", ""),
|
341 |
+
"arxiv_link": row.get("arxiv_link", ""),
|
342 |
+
"publication_date": row.get("publication_date", ""),
|
343 |
+
"raw_latex_related_works": row.get("raw_latex_related_works", ""),
|
344 |
+
"clean_latex_related_works": row.get("clean_latex_related_works", ""),
|
345 |
+
"pdf_related_works": row.get("pdf_related_works", ""),
|
346 |
+
}
|
347 |
+
|
348 |
+
elif split == "citations":
|
349 |
+
with open(filepath, encoding="utf-8") as f:
|
350 |
+
reader = csv.DictReader(f)
|
351 |
+
for key, row in enumerate(reader):
|
352 |
+
yield key, {
|
353 |
+
"parent_paper_title": row.get("parent_paper_title", ""),
|
354 |
+
"parent_paper_arxiv_id": row.get("parent_paper_arxiv_id", ""),
|
355 |
+
"citation_shorthand": row.get("citation_shorthand", ""),
|
356 |
+
"raw_citation_text": row.get("raw_citation_text", ""),
|
357 |
+
"cited_paper_title": row.get("cited_paper_title", ""),
|
358 |
+
"cited_paper_arxiv_link": row.get("cited_paper_arxiv_link", ""),
|
359 |
+
"cited_paper_abstract": row.get("cited_paper_abstract", ""),
|
360 |
+
"has_metadata": _safe_bool_convert(row.get("has_metadata", "False")),
|
361 |
+
"is_arxiv_paper": _safe_bool_convert(row.get("is_arxiv_paper", "False")),
|
362 |
+
"bib_paper_authors": row.get("bib_paper_authors", ""),
|
363 |
+
"bib_paper_year": _safe_float_convert(row.get("bib_paper_year", "")),
|
364 |
+
"bib_paper_month": row.get("bib_paper_month", ""),
|
365 |
+
"bib_paper_url": row.get("bib_paper_url", ""),
|
366 |
+
"bib_paper_doi": row.get("bib_paper_doi", ""),
|
367 |
+
"bib_paper_journal": row.get("bib_paper_journal", ""),
|
368 |
+
"original_title": row.get("original_title", ""),
|
369 |
+
"search_res_title": row.get("search_res_title", ""),
|
370 |
+
"search_res_url": row.get("search_res_url", ""),
|
371 |
+
"search_res_content": row.get("search_res_content", ""),
|
372 |
+
}
|
373 |
+
elif split == "important_citations":
|
374 |
+
with open(filepath, encoding="utf-8") as f:
|
375 |
+
reader = csv.DictReader(f)
|
376 |
+
for key, row in enumerate(reader):
|
377 |
+
yield key, {
|
378 |
+
"parent_paper_title": row.get("parent_paper_title", ""),
|
379 |
+
"parent_paper_arxiv_id": row.get("parent_paper_arxiv_id", ""),
|
380 |
+
"citation_shorthand": row.get("citation_shorthand", ""),
|
381 |
+
"raw_citation_text": row.get("raw_citation_text", ""),
|
382 |
+
"cited_paper_title": row.get("cited_paper_title", ""),
|
383 |
+
"cited_paper_arxiv_link": row.get("cited_paper_arxiv_link", ""),
|
384 |
+
"cited_paper_abstract": row.get("cited_paper_abstract", ""),
|
385 |
+
"has_metadata": _safe_bool_convert(row.get("has_metadata", "False")),
|
386 |
+
"is_arxiv_paper": _safe_bool_convert(row.get("is_arxiv_paper", "False")),
|
387 |
+
"bib_paper_authors": row.get("bib_paper_authors", ""),
|
388 |
+
"bib_paper_year": _safe_float_convert(row.get("bib_paper_year", "")),
|
389 |
+
"bib_paper_month": row.get("bib_paper_month", ""),
|
390 |
+
"bib_paper_url": row.get("bib_paper_url", ""),
|
391 |
+
"bib_paper_doi": row.get("bib_paper_doi", ""),
|
392 |
+
"bib_paper_journal": row.get("bib_paper_journal", ""),
|
393 |
+
"original_title": row.get("original_title", ""),
|
394 |
+
"search_res_title": row.get("search_res_title", ""),
|
395 |
+
"search_res_url": row.get("search_res_url", ""),
|
396 |
+
"search_res_content": row.get("search_res_content", ""),
|
397 |
+
}
|
papers.csv
DELETED
The diff for this file is too large to render.
See raw diff
|
|
papers_with_related_works.csv
ADDED
The diff for this file is too large to render.
See raw diff
|
|
usage_example.py
ADDED
@@ -0,0 +1,175 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
"""
|
3 |
+
Usage example for the Lotus Deep Research dataset.
|
4 |
+
|
5 |
+
This shows how to use the dataset builder directly, which is the recommended approach
|
6 |
+
for local development and testing.
|
7 |
+
"""
|
8 |
+
|
9 |
+
import sys
|
10 |
+
import os
|
11 |
+
from pathlib import Path
|
12 |
+
|
13 |
+
# Add the current directory to Python path
|
14 |
+
sys.path.insert(0, str(Path(__file__).parent))
|
15 |
+
|
16 |
+
from lotus_deep_research import LotusDeepResearch
|
17 |
+
import pandas as pd
|
18 |
+
|
19 |
+
|
20 |
+
def load_papers_dataset():
|
21 |
+
"""Load the papers dataset."""
|
22 |
+
print("Loading papers dataset...")
|
23 |
+
|
24 |
+
# Create dataset builder
|
25 |
+
builder = LotusDeepResearch(config_name="papers")
|
26 |
+
|
27 |
+
# Mock download manager for local files
|
28 |
+
class MockDownloadManager:
|
29 |
+
def download_and_extract(self, url_or_path):
|
30 |
+
return url_or_path
|
31 |
+
|
32 |
+
dl_manager = MockDownloadManager()
|
33 |
+
|
34 |
+
# Get split generators and load data
|
35 |
+
split_generators = builder._split_generators(dl_manager)
|
36 |
+
split_gen = split_generators[0] # Get train split
|
37 |
+
|
38 |
+
# Collect all examples
|
39 |
+
papers_data = []
|
40 |
+
for key, example in builder._generate_examples(
|
41 |
+
split_gen.gen_kwargs["filepath"],
|
42 |
+
split_gen.gen_kwargs["split"]
|
43 |
+
):
|
44 |
+
papers_data.append(example)
|
45 |
+
|
46 |
+
print(f"Loaded {len(papers_data)} papers")
|
47 |
+
return papers_data
|
48 |
+
|
49 |
+
|
50 |
+
def load_citations_dataset():
|
51 |
+
"""Load the citations dataset."""
|
52 |
+
print("Loading citations dataset...")
|
53 |
+
|
54 |
+
# Create dataset builder
|
55 |
+
builder = LotusDeepResearch(config_name="citations")
|
56 |
+
|
57 |
+
# Mock download manager for local files
|
58 |
+
class MockDownloadManager:
|
59 |
+
def download_and_extract(self, url_or_path):
|
60 |
+
return url_or_path
|
61 |
+
|
62 |
+
dl_manager = MockDownloadManager()
|
63 |
+
|
64 |
+
# Get split generators and load data
|
65 |
+
split_generators = builder._split_generators(dl_manager)
|
66 |
+
split_gen = split_generators[0] # Get train split
|
67 |
+
|
68 |
+
# Collect all examples
|
69 |
+
citations_data = []
|
70 |
+
for key, example in builder._generate_examples(
|
71 |
+
split_gen.gen_kwargs["filepath"],
|
72 |
+
split_gen.gen_kwargs["split"]
|
73 |
+
):
|
74 |
+
citations_data.append(example)
|
75 |
+
|
76 |
+
print(f"Loaded {len(citations_data)} citations")
|
77 |
+
return citations_data
|
78 |
+
|
79 |
+
|
80 |
+
def load_important_citations_dataset():
|
81 |
+
"""Load the important citations dataset."""
|
82 |
+
print("Loading important citations dataset...")
|
83 |
+
|
84 |
+
# Create dataset builder
|
85 |
+
builder = LotusDeepResearch(config_name="important_citations")
|
86 |
+
|
87 |
+
# Mock download manager for local files
|
88 |
+
class MockDownloadManager:
|
89 |
+
def download_and_extract(self, url_or_path):
|
90 |
+
return url_or_path
|
91 |
+
|
92 |
+
dl_manager = MockDownloadManager()
|
93 |
+
|
94 |
+
# Get split generators and load data
|
95 |
+
split_generators = builder._split_generators(dl_manager)
|
96 |
+
split_gen = split_generators[0] # Get train split
|
97 |
+
|
98 |
+
# Collect all examples
|
99 |
+
important_citations_data = []
|
100 |
+
for key, example in builder._generate_examples(
|
101 |
+
split_gen.gen_kwargs["filepath"],
|
102 |
+
split_gen.gen_kwargs["split"]
|
103 |
+
):
|
104 |
+
important_citations_data.append(example)
|
105 |
+
|
106 |
+
print(f"Loaded {len(important_citations_data)} important citations")
|
107 |
+
return important_citations_data
|
108 |
+
|
109 |
+
|
110 |
+
def main():
|
111 |
+
"""Main example function."""
|
112 |
+
print("Lotus Deep Research Dataset - Usage Example")
|
113 |
+
print("=" * 50)
|
114 |
+
|
115 |
+
# Load datasets
|
116 |
+
papers = load_papers_dataset()
|
117 |
+
citations = load_citations_dataset()
|
118 |
+
important_citations = load_important_citations_dataset()
|
119 |
+
|
120 |
+
print("\n📊 Dataset Statistics:")
|
121 |
+
print(f" - Papers: {len(papers)}")
|
122 |
+
print(f" - Citations: {len(citations)}")
|
123 |
+
print(f" - Important Citations: {len(important_citations)}")
|
124 |
+
|
125 |
+
# Show sample paper
|
126 |
+
if papers:
|
127 |
+
sample_paper = papers[0]
|
128 |
+
print(f"\n📄 Sample Paper:")
|
129 |
+
print(f" - Title: {sample_paper['title']}")
|
130 |
+
print(f" - ArXiv ID: {sample_paper['arxiv_id']}")
|
131 |
+
print(f" - Authors: {sample_paper['authors'][:100]}...")
|
132 |
+
print(f" - Abstract: {sample_paper['abstract'][:200]}...")
|
133 |
+
|
134 |
+
# Show sample citation
|
135 |
+
if citations:
|
136 |
+
sample_citation = citations[0]
|
137 |
+
print(f"\n📚 Sample Citation:")
|
138 |
+
print(f" - Parent Paper: {sample_citation['parent_paper_title']}")
|
139 |
+
print(f" - Cited Paper: {sample_citation['cited_paper_title']}")
|
140 |
+
print(f" - Has Metadata: {sample_citation['has_metadata']}")
|
141 |
+
print(f" - Is ArXiv Paper: {sample_citation['is_arxiv_paper']}")
|
142 |
+
|
143 |
+
# Show sample important citation
|
144 |
+
if important_citations:
|
145 |
+
sample_important_citation = important_citations[0]
|
146 |
+
print(f"\n⭐ Sample Important Citation:")
|
147 |
+
print(f" - Parent Paper: {sample_important_citation['parent_paper_title']}")
|
148 |
+
print(f" - Cited Paper: {sample_important_citation['cited_paper_title']}")
|
149 |
+
print(f" - Has Metadata: {sample_important_citation['has_metadata']}")
|
150 |
+
print(f" - Is ArXiv Paper: {sample_important_citation['is_arxiv_paper']}")
|
151 |
+
|
152 |
+
# Convert to pandas for easier analysis
|
153 |
+
print(f"\n🐼 Converting to Pandas DataFrames...")
|
154 |
+
papers_df = pd.DataFrame(papers)
|
155 |
+
citations_df = pd.DataFrame(citations)
|
156 |
+
important_citations_df = pd.DataFrame(important_citations)
|
157 |
+
|
158 |
+
print(f" - Papers DataFrame: {papers_df.shape}")
|
159 |
+
print(f" - Citations DataFrame: {citations_df.shape}")
|
160 |
+
print(f" - Important Citations DataFrame: {important_citations_df.shape}")
|
161 |
+
|
162 |
+
# Show some analysis
|
163 |
+
print(f"\n📈 Quick Analysis:")
|
164 |
+
print(f" - Unique parent papers in citations: {citations_df['parent_paper_arxiv_id'].nunique()}")
|
165 |
+
print(f" - Citations with metadata: {citations_df['has_metadata'].sum()}")
|
166 |
+
print(f" - ArXiv citations: {citations_df['is_arxiv_paper'].sum()}")
|
167 |
+
print(f" - Unique parent papers in important citations: {important_citations_df['parent_paper_arxiv_id'].nunique()}")
|
168 |
+
print(f" - Important citations with metadata: {important_citations_df['has_metadata'].sum()}")
|
169 |
+
print(f" - ArXiv important citations: {important_citations_df['is_arxiv_paper'].sum()}")
|
170 |
+
|
171 |
+
return papers_df, citations_df, important_citations_df
|
172 |
+
|
173 |
+
|
174 |
+
if __name__ == "__main__":
|
175 |
+
papers_df, citations_df, important_citations_df = main()
|