natcomSS / README.md
ZexiK's picture
Upload README.md with huggingface_hub
c1cb64a verified
# Nature Communications Articles Dataset - natcomm_2512_001
## Dataset Description
This dataset contains scraped Nature Communications articles with comprehensive extraction of:
- Main figures with captions and citations
- Tables with full HTML and data
- Supplementary figures with citation contexts
- Supplementary tables with citation contexts
- Supplementary files (PDFs, Excel files)
- Article sections with hierarchical structure (h2=section, h3=subsection)
- Figure/table citation contexts linking to sections
## Dataset Structure
### Files
- `images_tables_natcomm_2512_001.zip`: Contains all article images and tables
- `files/[article_id]/images/`: Main figure images
- `files/[article_id]/tables/`: Table HTML files
- `metadata_natcomm_2512_001.zip`: Contains JSON metadata for each article
- `metadata/[article_id].json`: Complete metadata including:
- `article_id`: Article identifier
- `url`: Original article URL
- `title`: Article title
- `image_set`: Dict of main figures with captions and citations
- `table_set`: Dict of tables with captions and data
- `supplementary_image_set`: Dict of supplementary figures with citations
- `supplementary_table_set`: Dict of supplementary tables with citations
- `supplementary_files`: List of supplementary file links
- `section_set`: Dict of article sections with hierarchical structure
## Statistics
- **Total Articles**: 1685
- **Scraping Date**: 2025-12-12
### Content Summary
- **Main Figures**: 8788 (5.2 per article)
- **Tables**: 689 (0.4 per article)
- **Supplementary Figures**: 13260 (7.9 per article)
- **Supplementary Tables**: 2795 (1.7 per article)
- **Supplementary Files**: 14588 (8.7 per article)
- **Sections**: 35630 (21.1 per article)
## Features
### Section Hierarchy
Each section includes:
- `level`: 2 (h2 main section) or 3 (h3 subsection)
- `type`: "section" or "subsection"
- `title`: Section heading
- `content`: Full section text
### Excluded Metadata Sections
Administrative sections are filtered out:
- Acknowledgements, Author information, Contributions
- Peer review, Ethics declarations, Rights and permissions
- Similar content recommendations
### Citation Tracking
Each figure/table has `contexts` array with:
- `section`: Section title where cited
- `text`: ±100 character context around citation
- `full_content`: Complete section content
## Usage
```python
import json
import zipfile
# Load article metadata
with open('metadata_natcomm_2512_001/metadata/s41467-025-65722-y.json', 'r') as f:
article = json.load(f)
# Access sections with hierarchy
for sec_id, sec_info in article['section_set'].items():
if sec_info['level'] == 2:
print(f"Section: {sec_info['title']}")
else:
print(f" Subsection: {sec_info['title']}")
# Access figures with citations
for fig_id, fig_info in article['image_set'].items():
print(f"{fig_id}: {fig_info['caption']}")
print(f" Cited in: {[c['section'] for c in fig_info['contexts']]}")
```
## Version Notes
**natcomm_2512_001** (December 2025):
- Fixed section extraction to exclude recommendation sections
- Added section hierarchy tracking (h2/h3 levels)
- Improved metadata section filtering
- Enhanced supplementary material citation tracking
## License
This dataset is derived from Nature Communications articles. Please respect original article licenses and cite appropriately.
## Created By
Batch Name: natcomm_2512_001
Scraping Tool: Nature Communications Comprehensive Scraper
Date: 2025-12-12
Repository: https://huggingface.co/datasets/ZexiK/natcomSS