README.md

---
license: mit
---

# ArteFact JSON Dataset

A structured metadata collection containing comprehensive information about artworks, creators, topics, and sentence-level annotations from the ArteFact art historical corpus.

## Dataset Overview

This dataset provides the foundational metadata and structured information that powers the ArteFact AI research platform. It contains 5 essential JSON files that enable cross-referencing between visual artworks, scholarly texts, and semantic embeddings.

## 🏗️ Data Structure

```markdown
artefact-json/
├── creators.json      # Artist and creator information
├── sentences.json     # Sentence-level annotations and metadata
├── topic_names.json   # Human-readable topic labels
├── topics.json        # Topic categorization and relationships
└── works.json         # Artwork metadata and descriptions
```

##  Dataset Statistics

- **Total Files**: 5 JSON files
- **Total Size**: ~1.2 GB
- **Data Coverage**: 7,200+ artworks and associated metadata

### File Details

| File | Size | Description | Records |
|------|------|-------------|---------|
| `creators.json` | 850 KB | Artist and creator information | Creator profiles |
| `sentences.json` | 1.18 GB | Sentence-level annotations | Sentence metadata |
| `topic_names.json` | 549 B | Human-readable topic labels | Topic mappings |
| `topics.json` | 1.85 MB | Topic categorization | Topic relationships |
| `works.json` | 8.5 MB | Artwork metadata | Work descriptions |

##  Data Schema

### `creators.json`
Contains biographical and professional information about artists, architects, and other creators referenced in the corpus.

### `sentences.json`
The largest file containing sentence-level metadata, annotations, and cross-references between textual content and visual artworks.

### `topic_names.json`
Human-readable labels for topic categories, enabling semantic search and filtering across the corpus.

### `topics.json`
Hierarchical topic relationships and categorization data for organizing and filtering art historical content.

### `works.json`
Comprehensive metadata about individual artworks, including titles, dates, materials, dimensions, and provenance information.

## 🚀 Usage

This dataset is designed to work in conjunction with:
- **`samwaugh/artefact-embeddings`** - Vector embeddings for semantic search
- **`samwaugh/artefact-markdown`** - Source scholarly texts and images

### Example Use Cases
- **Metadata-driven search** and filtering
- **Cross-referencing** between artworks and scholarly texts
- **Topic modeling** and categorization analysis
- **Creator relationship** mapping and analysis
- **Art historical research** and data mining
- **AI training** for art understanding models

##  Data Relationships

The JSON files are interconnected through shared identifiers:
- **Work IDs** (e.g., `W1009740230`) link works to their metadata
- **Creator IDs** connect artists to their works and biographical data
- **Topic IDs** enable categorization and semantic organization
- **Sentence IDs** provide granular access to textual content

## 📖 Citation

If you use this dataset in your research, please cite:

```bibtex
@dataset{artefact_json_2024,
  title={ArteFact JSON Dataset},
  author={Waugh, Samuel},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/samwaugh/artefact-json}
}
```

## 🔗 Related Resources

- **ArteFact Platform**: [samwaugh/artefact](https://huggingface.co/spaces/samwaugh/artefact)
- **Embeddings Dataset**: [samwaugh/artefact-embeddings](https://huggingface.co/datasets/samwaugh/artefact-embeddings)
- **Markdown Dataset**: [samwaugh/artefact-markdown](https://huggingface.co/datasets/samwaugh/artefact-markdown)

## 📄 License

This dataset is released under the MIT License. See the [LICENSE](LICENSE) file for details.

## Contributing

This dataset is part of the larger ArteFact project. For questions, issues, or contributions, please refer to the main ArteFact repository.

---

*Part of the ArteFact AI Research Platform - Bridging Visual Art and Textual Scholarship*