The SMarT Classifier for Arabic Fine-Grained Dialect Identification

Abstract

This paper describes the approach adopted by the SMarT research group to build a dialect identification system in the framework of the Madar shared task on Arabic fine-grained dialect identification. We experimented several approaches, but we finally decided to use a Multinomial Naïve Bayes classifier based on word and character ngrams in addition to the language model probabilities. We achieved a score of 67.73% in terms of Macro accuracy and a macro-averaged F1-score of 67.31%.

Machine Translation of Omani Arabic Dialect from Social Media

“…The PADIC (Parallel Arabic DIalect Corpus) (Meftouh et al, 2015) multi-dialectal corpus contains six dialects in addition to MSA. Two Algerian dialect corpora were created: Annaba's dialect (a city in Algeria) from daily conversations and the dialect from movies/TV shows in the Algiers dialect.…”

NADI 2023: The Fourth Nuanced Arabic Dialect Identification Shared Task

“…This prompted researchers to create new DA datasets, usually targeting a limited number of specific regions or countries (Gadalla et al, 1997;Diab et al, 2010;Al-Sabbagh and Girju, 2012;Sadat et al, 2014;Harrat et al, 2014;Jarrar et al, 2016;Khalifa et al, 2016;Al-Twairesh et al, 2018;Alsarsour et al, 2018;Kwaik et al, 2018;El-Haj, 2020). This was followed by several works that introduced multi-dialectal datasets and models for regionlevel dialect identification (Zaidan and Callison-Burch, 2011;Bouamor et al, 2014;Meftouh et al, 2015). The initial Arabic dialect identification shared tasks were part of the VarDial workshop series, primarily utilizing transcriptions of speech broadcasts (Malmasi et al, 2016).…”

Arabic Dialect Identification under Scrutiny: Limitations of Single-label Classification

“…(2) Translation in which participants are asked to translate sentences into their native Arabic dialects (Ho, 2006-;Meftouh et al, 2015;Bouamor et al, 2018;Mubarak, 2018). If all the participants are asked to translate the same source sentences, then the dataset is composed of parallel sentences in various dialects.…”

“…MPCA -/ 5 / 3 -2,000 Egyptian Arabic sentences from a pre-existing corpus, manually translated into 4 other country-level dialects in addition to MSA. PADIC (Meftouh et al, 2015) 5 / 4 / 2 -6,400 sentences sampled from the transcripts of recorded conversations and movie/TV shows in Algerian Arabic and manually translated into 4 other dialects and MSA. DIAL2MSA (Mubarak, 2018) -/ -/ 4 -Dialectal tweets manually translated into MSA.…”