Week 1 — Identifying Book Genre and Recommendation
If you are a regular reader after a while, it becomes difficult to find books that are similar to your favorite style. With this project, we aim to help the people which are interested in books like we are.
What attracts people to buy books is the visuality as well as the titles and description. If a suitable cover has been chosen for the book genre, it will both give the reader a better idea and attract more attention, because the book covers have artistic meanings both visually and literally. Therefore, the value that the book cover adds to the literature should be taken into consideration.
When we think about all of this, we decided to build a machine learning model by combining three factors: cover image, title, and description (optional). This model will classify the books by their genre. After the classification, a new book will be recommended in the same genre according to the features of the input book. To provide the user an ease of use in the application, we will set the description feature as optional and our model will work with the other two features too because generally, book descriptions take time to write down.
To realize this project we investigated some research papers and similar applications. For the cover image input, a group of researchers used the concept of transfer learning and developed a CNN-based system for book cover genre classification. AlexNet pre-trained on ImageNet is adapted for the task of genre recognition. For the text data like title and description, usually, algorithms like bag-of-words, or more advanced structures like RNN are used, and for the recommendation part, generally, similarity measurements are used. You can find the related works at the end of this post.
See you next week!
