A semantic-Based Model to Assess Information for Intelligence


J. Besombes, L. Cholvy, V. Dragos

PDF icon AL04-07.pdf1.38 MB

This paper addresses the problem of information evaluation for Intelligence. Starting from NATO recommendations for assessing information, we propose a semanticbased model to evaluate information. We also define a semi-automatic evaluation process in which an ontology is used, to detect similar items of information. Semantically similar items are then presented to an operator in charge, to estimate their correlations. Items of information are electronic documents exploited for intelligence purposes and they are  considered in a broader sense with respect to their form (structured files or free-form text) and their content (description of images of video scenes,  HUMINT reports). Linguistic variance, an inherent feature of textual data, can also be handled by using the ontology, while human intervention during the evaluation process ensures a good quality outcome. Finally, we show that this process is compliant with NATO recommendations, while going beyond their limitations.