Proposing Contextually Relevant Quotes for Images

Shivali Goel, Rishi Madhok, Shweta Garg

Abstract: Due to the rise in deep learning techniques used for the task of automatic image captioning, it is now possible to generate natural language descriptions of images and their regions. However, these captions are often too plain and simple. Most users on social media and other micro blogging websites use flowery language and quote like captions to describe the pictures they post online. We propose an algorithm that uses a combination of deep learning and natural language processing techniques to provide contextually relevant quotes for any given input image. We also present a new dataset, QUOTES500K, with the goal of advancing research requiring large dataset of quotes. Our dataset contains five hundred thousand (500K) quotes along with the author name and their category tags.