Applied text mining for military intelligence necessities

Natural language texts contain information that is not directly suitable for processing by a computer. However, the amount of texts to be analyzed often exceeds the capabilities of human reconnaissance specialists. Current intelligence, therefore, relies on the computational pre-processing of gathered HUMINT reports, SIGINT data and open sources in order to reduce the number of reports to be evaluated by human experts. State-of-the-art text mining featuring information extraction and emerging semantic role labeling can be used for automated threat recognition and offers relief from searching through the haystack [1]. In this paper we present a system tailored to the necessities of military text mining over HUMINT reports. We discuss the modules of the system to meet the needs of the application by armed forces. As a conclusion we show how that system can be used in the text mining process of automatic threat recognition.