Design and Development of Multilingual Sentiment Analysis Method using Aya-Expanse with Linear Transformer

Abstract

Sentiment analysis is an important method used to understand people’s opinions, emotions, and views from text data. Because social media sites like Twitter are so widely used, a lot of content is produced in a variety of languages, which presents difficulties for utilizing Natural Language Processing (NLP). Traditional methods often find it difficult to handle multilingual data because languages have different grammar, structure, and cultural expressions. Also, common deep learning models can be slow and inefficient when processing large amounts of real-time data. It makes it harder to use them for quick sentiment tracking on fast-changing social media platforms. To address these concerns, an efficient multilingual sentiment analysis is suggested on Twitter data in this article. Initially, from the available datasets, the required multilingual text is garnered. Further, the garnered multilingual text is subjected to a hybrid multilingual sentiment analysis framework named Aya-Expanse with Linear Transformer (AE-LT) for performing sentiment analysis. Here, the AE produce the high-quality vector representations of multilingual text, which is processed by the LT for accurate sentiment analysis. To confirm the superiority of the proposed model, its performance is finally compared with baseline models.