Video Mining: Measuring Visual Information Using Automatic Methods


SpeakerDr Xin WANG

Site: B129

Time 930-11:30   Tuesday Nov.14 2017

Abstract: Marketers are becoming increasingly dependent on videos to market their products and services. However, there is no standard set of measures of visual information that can be applied to large datasets. In this article, the authors propose three standard measures that are automatically obtained from videos: video duration, visual variation and video content. The authors test the measures on advertising videos from Kickstarter, an online crowdfunding website, and show that all these measures have significant effects on the funding outcomes of the projects. The research benefits both academic and practitioners. Instead of the expensive controlled lab experiments, these measures can be effectively implemented and used for large datasets. Researchers could apply these measures of other sets of visual information. Marketers could use the research to guide their video design and improve their advertising effectiveness.