The internet and its increasing access have significantly impacted contemporary society, with mental health becoming a significant issue due to the Covid-19 pandemic and social segregation standards. As a result, online platforms, particularly social media and video calling services, WhatasApp chat have become the primary means of communication. AI is essential to these technologies, and as we go toward automation and remote computing, the relationship between these technologies and human users becomes ever more important. Sentiment analysis is a technique that classifies opinions in online data by utilizing Natural Language Processing (NLP), Machine Learning (ML), and Statistics. It can increase user happiness, support mental health, spot dangers on social media, and help parents evaluate the mental health of their kids. In this context, the authors have analyzed the Multi-Class Mental State Decoder using a specific model based on the BERT Algorithm. The procedure can be used for many different things, such as tracking social media comments, improving consumer happiness, and spotting dangers. In order to improve the final feature vector accuracy for the sentiment classification results and better blend the context for identifying sentiment, the BERT Transformer model with BiLSTM constructs a comparable sentiment analysis model with respect to the consumer review content. The method presented in this research was compared with three other methods that used the same data set and was simulated through trials. The suggested approach has the highest precision, recall, and F1-Measure, with values of 92.64%, 90.32%, and 98.62%, respectively, according to the data.