Intelligent Book Recommender System Using Machine Learning and NLP

In an age where digital literature is growing at an incredible pace, helping readers discover meaningful and relevant books is more important than ever. Traditional recommendation systems, focused on user history and explicit feedback, often fail to capture the deeper meaning of a book’s content—especially in cold-start scenarios where limited user data is available. This paper presents a smart and scalable book recommen- dation system based on Sentence-BERT (SBERT), a trans- former model designed to understand sentence-level meaning. By converting book descriptions and user inputs into semantic embeddings, and comparing them using cosine similarity, the system is able to deliver accurate and contextually relevant book suggestions. We evaluated the model on large, real-world datasets such as Goodreads, Amazon Books, and EdNet. Across these datasets, our approach consistently outperformed traditional collaborative filtering and neural models in key metrics like Precision@10, Recall@10, and MRR. Because the system requires no prior user interaction data, it is ideal for digital libraries, e-learning platforms, and online bookstores. Its real-time performance and easy integration with modern APIs make it practical for wide deployment, and open the door to more personalized reader experiences.