Computational Linguistic Approach to Orthographic Representation of Egyptian Arabic: Challenges and Implications

In the past, Arabic Dialects (AD) have been poorly documented linguistically due to the lack of written forms and orthographies. However, in recent years, AD have become more widely used as a means of communication since social media and the everywhere availability of the internet have created a massive overflow of information and textual data, leading to a growing interest in Natural Language Processing (NLP) for these dialects. The highly inflectional morphology and the lack of standard orthography for AD pose an important challenge for NLP work. In this article, we handle the problem of lacking standard orthography during our work to build a morphological analyzer for Egyptian Arabic (EGY). To identify the guidelines for detecting conventional orthography, we depend on a corpus of 597,000 words that were gathered from various sources and genres. While analyzing the corpus morphologically, we handle the conventional orthography problem by assigning each word the conventional EGY Lemma and stem as close as possible to the EGY pronunciation no matter how it is typically written. Nevertheless, there are some common phenomena and complex cases involved in detecting conventional orthography during the morphological annotation process. Therefore, we take a closer look at and discuss these common phenomena and complex cases as we detect conventional orthography. These conventional orthographies are represented in a manner that facilitates the parsing of them correctly by the morphological analyzer. We tested the coverage of our morphological analyzer and compared it to one of the state-of-the-art morphological analyzers.