Sentiment analysis of recent flood disaster of Bangladesh from social media Bangla comments using deep learning

Abstract

This study provides complete sentiment analysis of Bengali social media posts made in the course of the country's recent flood disaster. We use deep learning models, such as CNN and Bi-LSTM, to evaluate a dataset of 4036 entries that have two attributes. Three groups of attitudes are identified by the target attribute: Fear,Neutral, and Religious. Our research uses advanced sentiment analysis tools to get an insight of the affected community's emotional reactions. Deep learning models like Bi-LSTM and CNN in particular are used to extract hidden expressions from the Bangla comments. The CNN model stands out as the most successful with a brilliant accuracy of 97.91%. This describes the model's strong capacity to identify emotions during a natural disaster and highlights its advantage over Bi-LSTM. The results provide insightful information about the range of emotional facets of the community's post-flood online discourse. The CNN model's success highlights the need for customized deep learning techniques for sentiment analysis within the particular context of social media information connected to disasters. This study adds to our understanding of how people feel during times of crisis and indicates how well deep learning models—in particular, CNN—work at identifying patterns in the Bangla social media comments made during the recent flood disaster in Bangladesh. Keywords: Sentiment