📄 Standardising the Arabic language and script: regional and global cultural stakes

Summary

This is a position paper, not an experimental study. It examines how Arabic language and script are coping with international standardisation in the digital era. The authors separate two layers: the writing system and the language itself. On the script side, they trace a long and fragmented history. Regional efforts failed to produce one Arabic encoding, and stability arrived only through Western and international bodies (ECMA, ISO 8859, Unicode, ISO 10,646). Some problems still remain, including bidirectional text, ligatures, hyphenation and localisation. Arabic also has a weak presence on the web and in domain names (ccTLD, iDN). On the language side, the authors argue that the main gap is now terminology. Arabic scientific and technical vocabulary is coined inconsistently and too slowly. They propose aligning Arabic term production with ISO TC37 methods, especially the concept-first approach of ISO 704. Their final recommendation is to put very large structured corpora online, both Arabic and bilingual, including speech and broadcast material. Such corpora would anchor terminology in real usage and support future machine translation.

Key points

• Early Arabic encoding attempts fragmented and no unified regional standard emerged:
  • ASMO's work and several CODAR versions (CODAR1, CODAR2, CODAR-U) never converged.
  • Talks ran from 1976 until ECMA proposed a code in February 1982.
  • By contrast, Chinese, Japanese and Korean users jointly pushed a CJK 16-bit coding.
• Unicode/ISO 10,646 largely settled character coding. Interface issues remain unresolved, including bidirectionality, ligatures and line breaking in multilingual text.
• An ITU report from 2011 estimated Arabic content at about 3% of the Internet.
• In the May 2011 ITU ranking, the best Arabic-country Latin ccTLD was .ma at 66th. It was followed by .sa (71st), .ly (79th), .jo (80th) and .tn (81st).
• Terminologists and lexicographers converged on the onomasiological method (concept → term) as the standard, codified in ISO 704. The paper also cites TEI and ISO TC37 as standards for describing language resources.

Interesting ideas

• I1. Language and writing system should be treated as distinct layers, since one script serves many languages and one language can change scripts.
• I2. Concept-first (onomasiological) terminology means you define the concept, then attach the terms that name it, instead of starting from words and guessing meanings.
• I3. Several competing local variants of one standard (the CODAR versions) delayed progress, while a single coordinated scheme (CJK) succeeded.
• I4. Terminology gains authority only when it is grounded in large, diverse usage corpora, including bilingual and spoken ones, not just in committee decisions.
• I5. Bilingual resources described with shared annotation standards (TEI, ISO TC37) are presented as the foundation for future translation.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tokens): Keep script and orthographic normalisation as an explicit layer separate from morphology and lemmas, so that encoding quirks never leak into lexical decisions.
• I2 (world): Key the ontology on concept IDs (WordNet/FrameNet/Wikidata) with surface terms as attached labels. This makes synonyms and cross-lingual names resolve to one explainable node.
• I3 (tree): When converting treebanks between UD dialects, normalise everything into one canonical scheme early, rather than maintaining several partial conversions.
• I4 (learn): Validate lexicon and verb-class entries against corpus attestation, and treat entries with no usage evidence as suspect during training and evaluation.
• I5 (back): Align bilingual texts at the meaning-graph level, so that translation through the meta-world can be checked against real parallel usage.