Understanding Public Sentiment on Social Media Platforms of Bangladesh: A Machine Learning Based Approach

Abstract

Sentiment analysis, also known as opinion mining, holds significant importance in today's digital age where vast amounts of textual data are generated daily. Understanding the sentiments expressed in text can provide valuable insights into public opinion, customer satisfaction, market trends, and social dynamics. While sentiment analysis has been extensively studied for major languages like English, there is a growing need for similar research in languages like Bangla. Bangla being the seventh most spoken language in the world, with a large number of users using social media to express their opinions and sentiments in this language. Analyzing sentiments in Bangla text allows for a deeper understanding of the sentiment landscape within the Bangla speaking community. Bangla language has its own unique cultural nuances, expressions, and sentiments that may not be captured accurately by models trained on other languages. Developing sentiment analysis models specifically for Bangla ensures cultural relevance and accurate interpretation of sentiments. Also little to none work has been done to successfully train any model on a Bangla dataset having more than 6 classes. Only Positive, Negative and Neutral classifications can’t portrait the exact sentiment of a sentence. So, we have built a custom Bangla Dataset which is classified into 6 classes of emotions: Joy, Sadness, Anger, Fear, Surprise and Disgust. Then we have trained different Neural Network and Machine Learning based models like: LSTM, CNN, BiLSTM, BiGRU and an Ensemble Learning Model on our dataset. Later, a comparative analysis based on the performance of the models is shown. The findings of this research contribute to sentiment analysis in Bangla language, shedding light on effective deep learning architectures for accurate emotion classification. The developed models can be utilized in various applications, such as social media sentiment analysis, customer feedback analysis, and opinion mining. The research also highlights the importance of considering linguistic and cultural nuances when designing sentiment analysis systems for specific languages