Prof. Abdelhadi Soudi: Standard Arabic-to-Moroccan Sign Language Machine Translation
DiTAWG: Digital Text Analysis Working Group
In this presentation, Prof. Abdelhadi Soudi will describe some challenges in an ongoing project on Standard Arabic (SA)-to-Moroccan Sign Language (MSL) machine translation. In Arabic, word structure is not built linearly as is the case in concatenative morphological systems, which results in a large space of of morphological variation. The language has a large degree of ambiguity in word senses, and further ambiguity attributable to a writing system that omits diacritics. (e.g. short vowels, consonant doubling, inflection marks). The lack of diacritics coupled with word order flexibility are causes of ambiguity in the syntactic structure of Arabic. The problem is compounded when translating into a visual/gestural language that has far fewer signs than words of the source language.
In this presentation, he will show the architecture of the system and provide a demo of several input examples with different levels of complexity. The Moroccan Sign Language database has currently 2000 Graphic signs and their corresponding video clips. The database extension is an ongoing process task that is done in collaboration with MSL interpreters, deaf signers and educators in Deaf schools in different regions in Morocco.