Sentiment Analysis on Multilingual Customer Reviews of Mobile Telecommunication Services Using Machine Learning and Deep Learning

Telecommunication networks, including the internet and various communication protocols, have created a global web of connectivity. Several telecommunication companies have emerged in Bangladesh. They are competing with each other to provide their users with superior experiences. This encompasses the interchange of information, concepts, preferences, and other modes of communication. With millions of customers interacting with these companies daily through various platforms, such as social media, surveys, customer support, and online reviews, gathering and going through millions of feedback would take a lot of time and work. The purpose of this study is to analyze the customer feedback on mobile telecommunication companies in Bangladesh using sentiment analysis. Due to the lack of labeled data, a large number of multilingual comments were extracted from several verified Facebook pages of telecommunication operators in Bangladesh and categorized to utilize. To evaluate the sentiment of the multilingual comments, machine learning algorithm, Support Vector Machines (SVM) was integrated. This study also included Long Short-Term Memory network (LSTM) and Bidirectional LSTM (BiLSTM) to tackle this discrepancy by harnessing the advanced capabilities of deep learning models. Subsequently, the performance of the models was assessed by obtaining potentially relevant measures, with the experimental results revealing the accuracy levels achieved across all algorithms and demonstrating their relative informativeness. This sentiment analysis will help telecommunication operators to gain a comprehensive understanding of client attitudes regarding new initiatives, products and offers, enabling companies to refine their services and remain competitive in this rapidly evolving and customer-centric sector.