Context Sensitive Dynamic Network Analysis to Support Military Missions
Vast quantities of unformatted texts are generated daily and need to be rapidly processed and encoded to understand, predict, explain, and manage behavior. Whether the focus is understanding changes in lED deployment and related activities, creating the most effective brigades, or enabling joint-force analysis of terrorist activities, this vast quantity of data needs to be processed, encoded, analyzed and interpreted. To do this suites of tools are needed. One key type of tool that is needed are dynamic network analysis tools (DNA) that can locate the networks embedded within and among these texts using content and meta-data; assess these networks, and locate critical patterns across groups and times.
Currently, there exists a wide array of network analytic and text mining tools; however, these tools do not meet the military need because: a) most do not scale well, b) they have not been linked together and work only in a standalone fashion, c) they require extensive effort to format the data to be analyzed, d) they cannot handle streaming information, e) they cannot handle missing data or data at varying levels of classification, and f) they provide only static analysis rather than enabling forecasting. As a result, vast quantities oftime are spent by the user in locating, ingesting and coding data and entering it in databases detracting from the time available for analyses and interpretation. Analysis are only rarely used as a bases for forecasting, and rarely are large numbers of forecasts considered as is possible with multi-agent simulations. What is needed are tools, and integration techniques, so that the time spent on data encoding is minimized, routine acts automated, and time available for analysis and interpretation increased. What is needed are tools and algorithms for taking extracting data and using it to instantiate models that enable forecasting. In principle, network techniques inform data extraction, analysis and prediction; however, current techniques are limited by their inability to handle context.