Why Microsoft Arabic Spell checker is ineffective

Alexis Amid Neme (1)

1 LIGM - Laboratoire d'Informatique Gaspard-Monge

Since 1997, the MS Arabic spell checker was integrated by Coltec-Egypt in the MS-Office suite and till now many Arabic users find it worthless. In this study, we show why the MS-spell checker fails to attract Arabic users. After spell-checking a document (10 pages -3300 words in Arabic), the assessment procedure spots 78 false positive errors. They reveal the lexical resource flaws: an unsystematic lexical coverage of the feminine and the broken plural of nouns and adjectives, and an arbitrary coverage of verbs and nouns with prefixed or suffixed particles. This unsystematic and arbitrary lexical coverage of the language resources pinpoints the absence of a clear definition of a lexical entry and an inadequate design of the related agglutination rules. Finally, this assessment reveals in general the failure of scientific and technological policies in big companies and in research institutions regarding Arabic.

Mots clés: Computational Linguistics; Arabic; NLP; spelling error detection; dictionary

Domaines: Informatique et langage [cs.CL]

Alexis Amid Neme. Why Microsoft Arabic Spell checker is ineffective. Linguistica Communicatio, 2014, Arabic Language in Information Technology, 16, pp.55. ⟨hal-01081965⟩