nltk-data-hub/crubadan

NLTK Crúbadán Language ID Corpus

Character 3-gram frequency tables for 449 writing systems, collected by Kevin Scannell's An Crúbadán web crawler (2010). Distributed via NLTK.

Trigrams use < (word start) and > (word end) as boundary markers.

Configs

Config	Description	Schema
table	Language metadata	crubadan_code, iso639_3, language_name
{lang_code}	Per-language trigrams	count, trigram

All 449 language codes: ab, abn, ace, ach, acu, ada, af, agr, aja, ak, ako, alt, amc, ame, am, ami, amr, an, ang, ar, … (and 429 more)

Schema

table

Column	Type	Description
crubadan_code	string	Internal Crúbadán writing-system code
iso639_3	string	ISO 639-3 language code
language_name	string	English language name

{lang_code} — one config per writing system

Column	Type	Description
count	int64	Frequency of trigram in crawled text
trigram	string	3-character sequence (</> = word boundaries)

Rows are sorted by descending count (most frequent first).

Sample languages

Code	ISO 639-3	Language
ab	abk	Abkhaz
abn	abn	Abua
ace	ace	Aceh
ach	ach	Acholi
acu	acu	Achuar-Shiwiar
ada	ada	Dangme
af	afr	Afrikaans
agr	agr	Aguaruna
aja	aja	Aja
ak	aka	Akan

Usage

from datasets import load_dataset

# Language metadata
meta = load_dataset("nltk-data-hub/crubadan", "table")
df = meta["crubadan"].to_pandas()

# Trigrams for a specific language
ds = load_dataset("nltk-data-hub/crubadan", "af")   # Afrikaans
trigrams = ds["crubadan"].to_pandas()               # count, trigram columns

Via NLTK

import nltk
nltk.download("crubadan")

reader = nltk.corpus.crubadan
reader.lang_codes()          # list all 449 codes
reader.trigrams("af")        # Afrikaans trigrams
reader.iso_lang_code("af")   # → 'afr'
reader.lang_name("af")       # → 'Afrikaans'

License

GPL v3 — © 2010 Kevin P. Scannell. See GNU GPL v3.

Citation

@inproceedings{crubadan,
  author    = {Scannell, Kevin P.},
  title     = {The Crúbadán Project: Corpus building for under-resourced languages},
  booktitle = {Building and Exploring Web Corpora: Proceedings of the 3rd Web as Corpus Workshop},
  year      = {2007},
  pages     = {5--15},
  url       = {http://borel.slu.edu/crubadan/}
}