Creating Parallel Arabic Dialect Corpus: Pitfalls to Avoid

Salima Harrat (1) , Karima Meftouh (2) , Kamel Smaïli (1)

1 SMarT - Statistical Machine Translation and Speech Modelization and Text
2 LRI-ANNABA - Laboratoire de Recherche en Informatique

Creating parallel corpora is a difficult issue that many researches try to deal with. In the context of under-resourced languages like Arabic dialects this issue is more complicated due to the nature of these spoken languages. In this paper, we share our experiment of creating a Parallel Corpus which contain several dialects and Modern Standard Arabic(MSA). We attempt to highlight the most important choices that we did and how good were these choices.

Mots clés: Parallel corpora; Modern Standard Arabic; Arabic dialects; Dialectal corpora

Domaines: Informatique et langage [cs.CL]

Salima Harrat, Karima Meftouh, Kamel Smaïli. Creating Parallel Arabic Dialect Corpus: Pitfalls to Avoid. 18th International Conference on Computational Linguistics and Intelligent Text Processing (CICLING), Apr 2017, Budapest, Hungary. ⟨hal-01557405⟩