Social Media Data Analysis - NATO Research T.DOCVIP

Social Media Data Analysis: A Dynamic Perspective Zhigang Wang, Ph.D. Department of National Defence Canada 101 Colonel By Drive Ottawa, Ontario K1A 0K2 Canada zhigang.wang@forces.gc.ca ABSTRACT Internet-based social media have changed the way that people communicate and interact with others. As more and more people engage in internet-based social media to share information and develop social relationships, the disruptive impact of internet-based social media on national defence and security need to be closely examined. In order to detect, minimize, and/or prevent the disruptive impact of social media, rigorous data analysis is critical. The emergence of internet-based social media challenges analytical techniques for data analysis. Compared to the data gathered through traditional means (e.g., experiments, surveys and interviews), social media data include much richer content (e.g., text messages, photos, videos, etc., generated by millions of users), and metadata (“data about data”). Given the richness and the high volume of social media data, appropriate analytical techniques are crucial for valid and accurate data assessments. The consequences of conducting data analyses with inappropriate analytical techniques are poor data assessments and incomplete information for senior military leaders. Currently, many analytical techniques such as content/semantic analysis and social network analysis have been adopted to evaluate some aspects of social media data. These techniques assist in detecting patterns in the content or the social relationships among social media users. These analytical techniques, however, are limited in exploring the important interactions among users, as well as the connections between content, the metadata, and users. As a result, a dynamic perspective on data analysis has emerged, which includes the adaptation of dynamic network analyses, dynamic graph algorithms, and visual analytics. This paper explains this dynamic perspective and its stra