Hugging Face's logo

Datasets:
lukeslp
/
language-data

Tasks:
Feature Extraction
Modalities:
Geospatial
Size:
100K<n<1M
Tags:
linguistics
languages
typology
phonology
glottolog
wals
License:
Dataset card Files
xet
 Community
language-data / README.md
lukeslp's picture
lukeslp
Upload README.md with huggingface_hub
d942b9d verified
preview code
|
raw
history blame contribute delete
4.02 kB
metadata 
license:
  - mit
  - cc-by-4.0
  - cc-by-sa-3.0
task_categories:
  - feature-extraction
tags:
  - linguistics
  - languages
  - typology
  - phonology
  - glottolog
  - wals
  - phoible
  - endangered-languages
  - language-families
  - geospatial
size_categories:
  - 100K<n<1M

Language Data

Curated linguistic datasets covering 23,740 languages. Pulls from Glottolog, WALS, PHOIBLE, and Joshua Project, plus original integration work combining them into a single queryable format.

Datasets

Core Files

File Records Size Source License
glottolog_coordinates.json 21,329 languoids 3.8 MB Glottolog CC-BY-4.0
glottolog_languoid.csv 23,740 languoids 1.8 MB Glottolog CC-BY-4.0
phoible.csv 105,484 phonemes 24 MB PHOIBLE CC-BY-SA-3.0
wals_languages.csv 3,573 languages 460 KB WALS CC-BY-4.0
wals_values.csv 76,475 feature values 4.5 MB WALS CC-BY-4.0
wals_parameters.csv 192 typological features 8 KB WALS CC-BY-4.0
world_languages_integrated.json 7,130 languages 8.1 MB Original integration MIT

Subdirectories

language-families/ -- Language family trees, Proto-Indo-European reconstructions, speaker counts. Restructured from Glottolog with enrichment from multiple sources. 10 JSON files covering family hierarchies up to 18 levels deep.

historical-corpora/ -- CLMET historical English corpus metadata.

lemmatization/ -- Historical lemma mappings (738 entries) for normalizing older English texts.

reference-corpora/ -- Brown Corpus part-of-speech statistics.

The Integrated Dataset

world_languages_integrated.json is the main original contribution. It merges four sources by ISO 639-3 code:

  • Glottolog -- coordinates, family classification, macroarea (7,040 languages)
  • Joshua Project -- speaker populations, demographics (7,130 languages)
  • Language history -- family tree positions, historical periods (198 languages)
  • US indigenous -- endangerment status, revitalization efforts (13 languages)
import json

with open("world_languages_integrated.json") as f:
    languages = json.load(f)

# Each record looks like:
# {
#   "iso_639_3": "eng",
#   "name": "English",
#   "glottolog": { "latitude": 53.0, "longitude": -1.0, "family_name": "Indo-European", ... },
#   "joshua_project": { "population": 379007140, "countries": [...], ... },
#   "speaker_count": { "count": 379007140, "source": "joshua_project_aggregated", ... },
#   "data_sources": ["joshua_project", "glottolog"]
# }

# Find endangered languages with coordinates
endangered = [
    lang for lang in languages
    if lang.get("glottolog", {}).get("latitude")
    and lang.get("speaker_count", {}).get("count", float("inf")) < 1000
]

This dataset is also available separately on HuggingFace as lukeslp/world-languages.

What's Not Here

ISO 639-3 code tables -- SIL International doesn't allow redistribution. Get them directly from iso639-3.sil.org.

License

Multiple licenses apply. See LICENSE.md for details.

  • Our work (integration, curation, language-families): MIT
  • Glottolog, WALS: CC-BY-4.0
  • PHOIBLE: CC-BY-SA-3.0 (share-alike)

Sources

Author

Luke Steuber