Abstract
HUMINT reports, as well as all other kinds of texts, can be pre-analyzed automatically in order to identify those items containing information about a specific problem of interest. Automated pre-analyzing tools often use deficient techniques of keyword spotting. However, automatic text analysis supported by linguistic principles offers techniques leading to more precise results. In this paper we present such a system tailored to the pre-analysis of HUMINT reports for automatic threat recognition. The paper discusses the components of our system and how they meet the requirements of the military domain. As a conclusion we show how indicators and their triggers can be defined in the system to tackle the problem of scattered information pieces and information overflow.