Judge a Book by its Cover: A Multimodal Approach to Book Genre Prediction

Authors

Reza Toosi, Alireza Hosseini, Ramin Toosi, Mohammad Ali Akhaee

Abstract

In today’s visually driven market, book cover design plays a crucial role in conveying a work’s narrative and thematic essence. However, conventional recommendation systems have largely overlooked the valuable information embedded in cover designs, particularly regarding genre classification. Recognizing a book’s genre from its cover is challenging due to the subtle and complex interplay of design elements. In this paper, we propose a multi-modal approach that integrates both visual and textual features extracted from book covers to predict their genres. Our method employs two complementary image embedding models—designed to capture both local and global visual information—alongside a text embedding module that incorporates cover text. We evaluated the proposed approach on a dataset comprising 57,000 books spanning 30 genres. Experimental results demonstrate that our method achieves a Top …