# Etymology Atlas: Global Language Relationships

4.17M+ etymological relationships across 19,400+ languages with geographic coordinates, phoneme inventories, and typological features.

## What's Inside

| File | Records | Description |
|------|---------|-------------|
| `etymologies.parquet` | 4.17M | Core etymology graph - word relationships with types and confidence |
| `languages.parquet` | 19,401 | Language metadata - families, coordinates, speakers, status |
| `cognate_sets.parquet` | 4,981 | Expert-annotated cognate groups from Lexibank (IE-CoR) |
| `phonemes.parquet` | 105,484 | PHOIBLE phoneme inventories with articulatory features |
| `linguistic_features.parquet` | 76,475 | WALS typological features (192 types) |

## Quick Start

```python
import pandas as pd

# Load main etymology table
etymologies = pd.read_parquet('etymologies.parquet')

# Find all cognates of "water"
water_etym = etymologies[etymologies['term1'].str.contains('water', case=False)]
print(f"Found {len(water_etym)} relationships")

# Get Indo-European languages
languages = pd.read_parquet('languages.parquet')
ie_langs = languages[languages['family_name'] == 'Indo-European']
print(f"{len(ie_langs)} Indo-European languages")
```

```r
library(arrow)

etymologies <- read_parquet('etymologies.parquet')
languages <- read_parquet('languages.parquet')

head(etymologies)
```

## Relationship Types

The `relationship_type` field includes 31 categories:

- **inherited** - Direct genetic descent (English "mother" ← Proto-Germanic)
- **borrowed** - Loanword from another language (English "café" ← French)
- **cognate** - Shared ancestry from common parent
- **derived** - Morphological derivation within a language
- **calque** - Loan translation (German "Wolkenkratzer" = "skyscraper")
- **semantic_shift** - Meaning change over time
- **compound** - Two roots combined
- **blend** - Portmanteau (brunch = breakfast + lunch)
- **back_formation** - Reverse derivation (edit ← editor)
- **clipping** - Shortened form (phone ← telephone)

Plus 21 more specialized types.

## Data Sources

| Source | License | Contribution |
|--------|---------|--------------|
| [etymology-db](https://github.com/droher/etymology-db) | CC BY-SA 3.0 | 3.75M etymology entries from Wiktionary |
| [Glottolog](https://glottolog.org/) | CC BY 4.0 | 19,401 languages with coordinates and families |
| [Lexibank IE-CoR](https://lexibank.clld.org/) | CC BY 4.0 | 4,981 expert cognate sets (Indo-European) |
| [PHOIBLE](https://phoible.org/) | CC BY 4.0 | 105,484 phoneme entries for 2,177 languages |
| [WALS](https://wals.info/) | CC BY 4.0 | 76,475 typological feature values |

## Research Ideas

1. **Etymology Networks** - Visualize word family trees and borrowing patterns
2. **Geographic Spread** - Map how words travel across language boundaries
3. **Sound Change** - Track phonological evolution using cognate sets + PHOIBLE
4. **Typological Correlations** - Find relationships between word order and other features
5. **Language Distance** - Build similarity metrics from cognate overlap
6. **Borrowing Patterns** - Which language families borrow most from each other?

## Schema Details

### etymologies.parquet

| Column | Type | Description |
|--------|------|-------------|
| term1 | string | Source word/morpheme |
| lang1 | string | ISO 639-3 code |
| lang1_name | string | Language name |
| lang1_family | string | Language family |
| term2 | string | Related word/morpheme |
| lang2 | string | ISO 639-3 code |
| lang2_name | string | Language name |
| lang2_family | string | Language family |
| relationship_type | string | One of 31 types |
| confidence | float | 0.0-1.0 (expert=1.0) |
| concept | string | Shared meaning (for cognates) |
| sources | string | Data provenance |

### languages.parquet

| Column | Type | Description |
|--------|------|-------------|
| glottocode | string | Glottolog identifier |
| iso_639_3 | string | ISO language code |
| name | string | Language name |
| family_name | string | Top-level family |
| family_glottocode | string | Family identifier |
| macroarea | string | Geographic region |
| latitude | float | Coordinate |
| longitude | float | Coordinate |
| status | string | living/endangered/extinct |
| speakers_count | int | Estimated speakers |
| phoneme_count | int | Number of phonemes |

## Citation

```bibtex
@dataset{etymology_atlas_2026,
  title = {Etymology Atlas: Global Language Relationships},
  author = {Steuber, Luke},
  year = {2026},
  publisher = {Kaggle},
  url = {https://kaggle.com/datasets/lukesteuber/etymology-atlas},
  license = {CC BY-SA 3.0}
}
```

## License

**CC BY-SA 3.0** - You can share and adapt this dataset if you:
- Give appropriate credit
- Distribute derivatives under the same license

Full text: https://creativecommons.org/licenses/by-sa/3.0/

## Updates

This dataset will be refreshed periodically as source data is corrected or expanded. Check the version number in metadata for the latest release.

## Related

- [Diachronica Etymology Explorer](https://diachronica.com/etymology/) - Interactive visualization
- [Lexibank](https://lexibank.clld.org/) - Source cognate datasets
- [Glottolog](https://glottolog.org/) - Language family reference