Arabic Natural Language Processing from Software Engineering to Complex Pipeline

Abstract

Arabic Natural Language Processing (ANLP) has known an important development during the last decade. Nowadays, Several ANLP tools are developed such as morphological analyzers, syntactic parsers, etc. These tools are characterized by their diversity in terms of development languages used, inputs/outputs manipulated, internal and external representations of results, etc. This is mainly due to the lack of models and standards that govern their implementations. This diversity does not favor interoperability between these tools or their reuse in new advanced projects. In this article, we propose APIs and models for three types of tools namely: stemmers, morphological analyzers and syntactic parsers, using SAFAR platform. Our proposal is a step for standardizing all aspects shared by tools of the same type. We review also the issue of interoperability between these tools. Finally, we discuss pipeline processes.