Title :
Towards Understanding the Effectiveness of Election Related Images in Social Media
Author :
Junhuan Zhu ; Jiebo Luo ; Quanzeng You ; Smith, J.R.
Author_Institution :
Univ. of Rochester, Rochester, NY, USA
Abstract :
In recent years, political campaigns have paid increasing attention to social media. During the election period, numerous election related images are posted. However, not all the images have the same effectiveness, and researchers have not investigated the intrinsic relationship between the effectiveness and the high-level visual features of social images. In this paper, we present a new study to analyze the effectiveness of election related images in social media. We first compute three semantic visual attributes for election related images: 1) face attribute, which indicates the presence of a political candidate, 2) text attribute, which describes the area of text information, 3) logo attribute, which denotes whether an image contains a campaign logo. Next, we consider the effectiveness in terms of the number of views and comments, and employ analysis of variance and association analysis to understand the importance of visual attributes in affecting the effectiveness of election related images. In addition, visual attributes distribution analysis reveals Obama campaign´s deliberate effort targeting social media. The experiments on the 2012 US presidential election related images provide interesting insight that can be exploited in similar scenarios.
Keywords :
feature extraction; government data processing; object detection; social networking (online); statistical analysis; Obama campaign; US presidential election; United States; analysis of variance; association analysis; campaign logo; election period; election related images; face attribute; high-level visual features; image understanding; logo attribute; political campaigns; political candidate; semantic visual attributes; social images; social media; text attribute; text information; Face; Face recognition; Image edge detection; Media; Nominations and elections; Semantics; Visualization; Effectiveness; Election; Social media;
Conference_Titel :
Data Mining Workshops (ICDMW), 2013 IEEE 13th International Conference on
Conference_Location :
Dallas, TX
Print_ISBN :
978-1-4799-3143-9
DOI :
10.1109/ICDMW.2013.112