<p>Recommendation systems, which offer recommendations for information automatically depending on the user, have grown in importance as web applications. Considering they give customers a better shopping experience, product recommendation systems are a significant part of the retail sector. Since retailers offer a wide range of products, through the identification of connections between products, recommendation systems provide the most effective way of showing customers to the products which are relevant to them. However, understanding the characteristics of consumers connected to various product relationships is equally important. It might be difficult to evaluate and understand the information found in reviews because they are frequently unstructured data. To gain more understanding of the emotions, opinions, and behaviors of a user, Sentiment analysis (SA) has achieves to a lot of attention as a solution to this problem. The accuracy and efficiency of these systems have increased significantly due to deep learning techniques. In this analysis, Sentiment analysis framework for entropy-based product recommendation system is presented. Amazon Kindle books dataset 2023 from Kaggle is used in this analysis. Entropy is used to reflect the global rating behavior of user and item and to provide diversity and novelty. Collaborative Filtering is combined with Sentiment Analysis to mitigate the cold-start issue. Bi-LSTM is used to process the text reviews, detect the contextual subtleties and sentiments in text. The combination of CF with SA and Bi-LSTM accurately recommends the books to users. By combining SA, CF and Bi-LSTM, proposed system improves the user satisfaction and personalized experience. F1-score, Root mean square error (RMSE), Mean Absolute Error (MAE), and accuracy are used to evaluate the performance of this model. Proposed model obtains 94.8% accuracy and 93% F1-score performance.</p>