Sentiment Analysis of Tunisian Users on Social Networks: Overcoming the Challenge of Multilingual Comments in the Tunisian Dialect

Analyzing Multilingual Conversations During COVID-19: An Imbalanced Class-Ensemble Learning Approach with Reweighted AdaBoost-SVM for Code-Switched Text Classification

“…In areas where bilingualism or multilingualism prevails, a notable trend is the merging of local languages with English or French for expressing opinions, a practice referred to as "code-mixing." This blending of languages in a single conversational thread is particularly evident in linguistically rich regions like Bangladesh [7], India [8], and Tunisia [9], the focus of our case study. In Tunisia, the prevalence of multilingualism on social networks can be attributed to a confluence of factors including historical colonization, the impacts of globalization and immigration, cultural openness, rising education levels, and the expansion of communication and international business [1], [9].…”

“…The COVID-19 pandemic has dramatically reshaped the landscape of social media communication, accelerating the trend towards multilingualism in online interactions. This phenomenon is particularly evident in the use of dialectical Arabic combined with code-switched Latin scripts, predominantly French and English, an emerging pattern in social media dialogue [1]. This combination of languages, while adding richness to digital exchanges, poses significant challenges in Text Classification, notably in Sentiment Analysis (SA).…”

“…The role of SA on social media has become increasingly pivotal, particularly within the healthcare sector amidst the Coronavirus pandemic. Such analysis is vital to grasp the public's response to the crisis, thereby informing decisions in managing the pandemic [1], [6]. This tool has become integral to the pandemic response strategies of various governments.…”

“…This blending of languages in a single conversational thread is particularly evident in linguistically rich regions like Bangladesh [7], India [8], and Tunisia [9], the focus of our case study. In Tunisia, the prevalence of multilingualism on social networks can be attributed to a confluence of factors including historical colonization, the impacts of globalization and immigration, cultural openness, rising education levels, and the expansion of communication and international business [1], [9]. In addition to code-mixing, a significant number of internet users exhibit a preference for Roman script or phonetic typing in their native language over Unicode, a trend known as "Transliteration."…”

“…A notable instance is the widespread use of Arabizi among Tunisian social media users. Arabizi, a unique linguistic blend that combines Arabic numerals with Roman script, has gained considerable traction in online communications across the Arab world [9]. Initially emerging as a practical solution for mobile communication and social media in the absence of Arabic-compatible interfaces, Arabizi has since evolved into a distinctive mode of expression in its own right.…”