Dialectal Machine Translation A Case Study

Author: Zohra LABED

Keywords: Machine translation, dialect, Arabic


Traditional grammarians are well-known for their clear
standing position against the spoken form of language. Primacy was, at their
time, orientated towards writing which was believed to represent, unlike
speech, the pure and correct version of language. Their attitude was previously
greatly influential. It was until the advent of structural linguistics that
language scholars started to claim the significance of studying the spoken
form. Speech was recognised as an autonomous object of investigation, but not
its variation. First linguists devoted their attention to stable linguistic
features and neglected those which vary. Variability meant for them disorder
and chaos. Language variation has however started to gain meditation since
sociolinguistics appeared. That language variation is systematic and worth
exploring has been realised gradually. Today, sociolinguists hold against any
kind of variable form marginalisation. Needless to say that the most salient
language variation is regional and lexical. One objective of sociolinguists is
to promote the spoken varieties or dialects, and provide them with the
necessary tools already supplied to the written form of language. Cooperation
between language scholars and computer scientists can turn automatic
translation possible to any dialectal variety. Given that dialects are
particularly exposed to lexical variation, the question which arises here: into
which item is the computer supposed to translate? Our choice has fallen on Oran
Arabic. In this paper, we will attempt to find an answer to this question on
fieldwork bases.