Book recommendation system using data mining techniques

In recent years, the growing number of published books has made it increasingly difficult for readers to select suitable titles. With the rise of e-commerce, access to a wide range of book genres has become significantly more convenient, much like with other consumer products. However, this accessibility has also heightened the importance of guiding readers toward appropriate book choices. E-commerce platforms often offer a much larger selection of books than physical bookstores, which can makes it harder for users during the decision-making process. One of the advantages of these platforms is the availability of user-generated reviews and evaluations, which assist in the selection process. This study aims to develop a content-based book recommendation system using data collected from the platform 1000kitap.com. A range of textual information related to books—particularly book summaries—was gathered and subjected to text preprocessing techniques. Subsequently, KeyBERT was employed to extract representative keywords from the summaries, and Zeyrek, a Turkish NLP library, was utilized to identify word stems. Furthermore, to enhance similarity computations, a deep learning model based on BERT was used to infer sub-genres within the novel category. The primary objective of this study is to calculate content-based similarity by comparing the extracted keywords of a user-selected book with those of other books in the dataset. Based on these similarity scores, the system recommends alternative books with related content, thereby assisting users in discovering new titles aligned with their interests.