ANALYZING IMAGE TWEETS IN MICROBLOGS

Abstract

Social media platforms now allow users to share images alongside their textual posts. These image tweets make up a fast-growing percentage of tweets, but have not been studied in depth unlike their text-only counterparts. Most existing studies on image tweets tackle tasks that are originated from text tweet domain, and their main effort is to incorporate generic image features (e.g., low-level features, deep learning features) to improve the performance of using text-only approaches.

In this thesis, we conduct a series of studies to answer four fundamental questions about image tweets: 1) What are the characteristics of image tweets? 2) What are the relationships between the image and text in image tweets? 3) How to model such image-text relationships? and 4) How to interpret the semantics of an image tweet? Answers to these questions will not only help us gain a deep understanding of image tweets and the related user behaviors, but also be beneficial to downstream applications.