Leveraging Book Covers to Develop a Book Recommendation Engine

University essay from KTH/Skolan för elektroteknik och datavetenskap (EECS)

Author: Hugo Carlsson; [2021]

Keywords: ;

Abstract: We propose a novel recommendation engine, capable of generating recommendations, whilst requiring limited user grades. The recommendation engine was developed together with the Swedish company Bokus, which is a bookstore with online and physical presence, offering more than 10 million different titles to its 2 million customers. In the report, we discuss how one can use state-of- the-art deep learning techniques to leverage book covers to generate book recommendations considering a limited number of axioms. Using book covers as the starting point for recommendations, solves the burdensome collection of user grades. From the findings in this report, we conclude that our convolutional neural network was able to generalize well over an unseen dataset, far exceeding the performance of random guesses. Furthermore, the proposed recommendation engine was able to generate meaningful recommendations, albeit, with varying relevance depending on genre. 

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)