Multilingual Sarcasm Detection Using Deep Learning and Transformed Based Model

Multilingualism refers to the ability to use three or more languages effectively, including using two or more languages by a person or community. Multilingual Sarcasm Detection identifies sarcasm in any language, differentiating it from non-sarcastic text. Sarcasm detection is challenging and requires advanced techniques. With the multilingualism of social media, Natural Language Processing (NLP) is required for text processing. NLP aims to develop emotion and sarcasm detection tools for several languages. Code-mixing and code-switching are issues. Multilingual speakers use local and global languages like English. Sarcasm detection is difficult in NLP as it requires more meaning. This study enhances sarcasm detection in Bangladeshi Bengali, English, and Banglish. The paper preprocesses languages to keep features in a combined dataset, using transformer classifiers and deep learning for multilingual classification. DistillBERT achieved 78% accuracy in cross-language sarcasm detection, enhancing low-resource language systems and sentiment analysis.