Sentiment analysis in Algerian dialects considering Code-switching

Abstract

Abstract : Code-switching is the act of using two or more languages in the same text due to cultural or contextual reasons. The code-switching property of text is an essential step to sentiment analysis, that piqued the interest of many researchers. Thus, the study of sentiment analysis when code-switching is involved have gained increased popularity, especially with the continuous rise of social media users and content that has more and more became integrated with our daily lives. The recent approaches used to deal with code-switching mostly depend on lexicons, and ready-to-use corpora, which makes it different and more difficult for the Arabic language, and its dialects, that use large and varied dialects and lexicons. In our research, we aim to address the problem of Arabic sentiment analysis when codeswitching is involved, and more precisely in the Algerian dialect. Many techniques have been used in previous researches including machine learning, and deep learning ones, in addition to getting aid form various NLP tools that were used to ease up the task. Our main objective is comparing various researches that has tried to deal with this challenge until today, by proposing different approaches to solve the problem at hand both in English and Arabic Languages using machine learning and deep learning. And after, we give a summary and compare between these methods.