Classification of Book Genres using Book Cover and Title

• Published in: 2019 IEEE International Conference on Intelligent Systems and Green Technology (ICISGT) pp. 72 - 723
• Main Authors: Biradar, Ganeshprasad R, JM, Raagini, Varier, Aravind, Sudhir, Manisha
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-06-2019
• Subjects: CNN; Computer architecture; Conferences; Feature extraction; GloVe; Image color analysis; Image Processing; Logistics; NLP; Pediatrics; Training; Xception
• DOI: 10.1109/ICISGT44072.2019.00031

The main purpose of this research was to determine if the cover and title of a book can be used to predict its genre. The main motivation behind this was that since the cover or title individually may not always be suggestive of the type of book, using both of them together may be more indicative of the genre of the book. The idea was to extract features from the cover and title and combine them to predict the genre of a book. A logistic regression model was used for training and prediction. An accuracy of 87.2% was obtained when we used the combined features. We got an accuracy of 67.04% using only cover features and an accuracy of 86.2% using only the title features. This shows that although using both the title and cover features gives better results, the title seems to more expressive of the genre of the book.