RESEARCH ON CO-CITATION NETWORK MAPPING
Abstract
N00014-21-S-B001 FY 2021Long Range BAA for Navy and Marine Corps Science and TechnologyAbstractGraphikas current technology identif ies large-scale online communities based on patterns of similarity intheir relationships and measures the propagation of information in real time across complex networks. Wehave recently begun an effort to further scale up our network analysis capabilities and pro vide automatedsignals of malign influence campaigns. As an extension to this effort, we propose creating a capabilitythat leverages Graphikas advanced network clustering methods in order to detect clusters of contentsources (domains, social influencers) and objec t_types (URLs, words, phrases, hashtags, visual content,etc.) that are propagated by similar actors, advancing on techniques of co-c itation analysis.Currently, Graphika clusters large network subgraphs based on patterns of relationships between subgraphnodes and s elected types of other relevant nodes, content sources, and content features. This processinvolves constructing a bipartite graph be tween subgraph nodes, designated as sources, and selectedtypes of nodes, content providers, and content features, designated colle ctively as targets. Currenttechnology clusters subgraph nodes based on patterns of co-attention to the same set of other nodes,con tent sources, and content features. The approach proposed here will apply the same clustering engineto the target mode of the bipart ite graph, thus flipping the axis, in order to discover clusters of nodes,content sources, and content features that are cited by similar actors.By clustering both actors and sources/content, we can seek to discover sets of sources that are amplifiedand cited in coordinated or structured ways, and we can begin to identify emerging narratives in socialnetworks, how different audiences engage with such narratives, and indicators of whether their spread isorganic or coordinated. It is hoped that this approach will produce a powerful method ofdimensionreduction for the analysis of large-scale content flows in online networks.By clustering both modes of t he author-network/content feature matrix, we will then observe patterns ofinteraction between network communities, and similarly pro pagated information streams, in order todetect both natural (sets of content features and sources with similar audience-interest sig natures) andcoordinated (information operations) phenomena.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 22, 2021
- Source ID
- N000142112956
Entities
People
- Vladimir Barash
Organizations
- Office of Naval Research
- United States Navy