Natural Language Processing Based Approach to Overcome Arabizi and Code Switching in Social Media Moroccan Dialect

Hajbi, Soufiane; Chihab, Younes; Ed-Dali, Rachid; Korchiyne, Redouan

doi:10.1007/978-3-030-91738-8_6

Cited by 4 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In 2023, Bousri et al [9] proposed a system for Rumor Detection in Algerian Franco-Arabic. In 2022, Hajbi et al [10] propose a new method based on NLP to address the challenges of pre-processing text that contains Franco-Arabic Based on Deep Learning and Associations. In 2021, Fourati et al [11] introduce a large Tunisian Arabizi dialectal dataset for Sentiment Analysis.…”

Section: Related Workmentioning

confidence: 99%

From Latin to Arabic: Innovative Approaches for Retrieving Original Texts and Vectorizing Franco-Arabic

Gabr¹,

Badr²,

Mahdi³

2023

INDJCSE

View full text Add to dashboard Cite

In the last few years, the importance of artificial intelligence, deep learning techniques, and transformerbased models in the analysis and understanding of English texts has emerged. It highlights this importance in various tasks, such as answering questions in smart chat systems, sentimental analysis, named entity recognition, and opinion polls analysis in various fields. The significant improvement in Internet services and their spread around the world caused a significant increase in the number of Internet users who speak Arabic. Interest in analyzing Arabic texts on various platforms has begun. This task posed a great challenge due to the difficulty of the Arabic language, its morphological richness, and the multiplicity of its dialects. In addition to the emergence of a new challenge, where many Arab users adopted writing the Arabic language using Latin letters due to the lack of support for the Arabic language at the beginning of the spread of the Internet. This way of writing is called the Franco-Arabic language or Arabizi. It is not English language nor Arabic language. We can call this process transliteration which concerns similar-sounding characters of another alphabet. Transliteration isn't always an exact science because sometimes words can be transliterated in more than one way. This caused uncertainty about converting Franco-Arabic to Arabic because there are many to many relationships between some Latin and Arabic characters in conversion based on the phonetic tone in addition to the multiplicity of writing methods from one Arab country to another. In this paper, we will introduce two methods for dealing with Franco-Arabic. The first method is concerned with retrieving original Arabic text based on pre-trained transformers-based models and the second method is concerned with training a new model that can vectorize and understand Egyptian Franco-Arabic texts directly.

show abstract

Section: Related Workmentioning

confidence: 99%