frontmatter 
title: "Arabic"
iso639_3: ara
iso639_1: ar
ran: "8/8/en-7/fr-7"
S: 8
M: 8
B_max: 7
ran_sum: 23
speakers: 422000000
tags: [language]

Arabic

8/8/en-7/fr-7

Arabic (ISO 639-3 ara). In words: hundreds of millions of fluent speakers; hundreds of millions of monolingual sentences; tens of millions of sentences parallel with English, tens of millions of sentences parallel with French.

Score breakdown

ComponentDigitRaw countReads asTop corpus
Speakers (S)8422,000,000hundreds of millions—
Monolingual (M)8672,096,859hundreds of millions—
Bilingual · English (en)787,893,588tens of millionsOpenSubtitles
Bilingual · French (fr)752,016,965tens of millionsOpenSubtitles

Benchmark performance

Published zero-shot results used to validate the notation (see the RAN Paper).

TaskBenchmarkMetricScore
MTNLLB-200 FLORESchrF++62.2
NERXTREME WikiANN XLM-R LargeF153.0
POSXTREME UD XLM-R LargeAccuracy67.5
Note

Macrolanguage; OPUS ar dominantly MSA + dialectal subtitles

Sources

  • Wikidata: Q13955
  • Counts use the sources this site applies consistently (see the overview) — Ethnologue (speakers), OSCAR 23.01 (monolingual), OPUS (bilingual) — except where this page's notes record a different source. The top corpus for each bilingual partner is shown in the breakdown above.

See the RAN overview for how scores are computed and what they mean.

Created · Updated