Advancing sentiment classification in Bangla text: an enhanced BERT approach on the SentNoB dataset

Abstract

Since social media has entered into our lives, it has set the scene for unmatched online communication and information sharing for a significant amount of time. These platforms enable people to share their thoughts, opinions and experiences, even in their native languages, which can be used as an asset for sentiment analysis. With this in mind, the paper conducts research about the application of Natural Language Processing (NLP) to evaluate and analyze sentiments from social media posts in Bangla language. In this present world, people tend to share their point of views across social medias over ongoing topics. The ample amount of personalized program data on numerous social media platforms presents an opportunity to gather large scales of information and to use non-invasive tools for sentiment analysis. The following research incorporates insights from twenty relevant studies, providing a clear image of existing methodologies and approaches. This study also acknowledges the challenges and opportunities that come with scooping out information from social media data, including issues of privacy, ethics, and data quality. Moreover, the research utilizes a combination of numerous NLP techniques, sentiment analysis and machine learning algorithms to embody robust models capable of identifying text sentiments accordingly. The proposed methodology’s observational assessment involves a large scale of social media databases which allows to assess the performance of models in real world aspects. The discovery illustrates the promising NLP-driven solutions in quick detection and in performing sentiment analysis.This study aims to detect users’ sentiment based on their posts posted in Bangla language with the help of BERT (BanglaBERT) and an ensemble algorithm (LSTM + BanglaBERT). This research highlights the potential of NLP-based approaches, specifically utilizing BERT in order to effectively identify and analyze mental health signals from social media posts in Bangla-offering a valuable tool for early intervention and mental health awareness.