Sentiment Analysis is key in NLP. It is very useful to understand the sentiments of the customers for the products and thus help the company to better improve the services or products from the feedback
BERT (Bidirectional Encoder Representation from Transformers) is useful for sentiment analysis. It is trained on large text corpus (Wikipedia + BooksCorpus) so it could understand the context very well and learn the embeddings of the words from training.
The dataset is obtained from Kaggle with 5 different sentiment ratings so it is a multi-classification problem. See details: Notebook