From Unstructured to Structured Information in Military Intelligence - Some Steps to Improve Information Fusion

摘要

Military commanders require timely and accurate understanding of the situation in their respective area of responsibility as well as a prediction of the likely intentions and capabilities of supposed or potential adversaries. To achieve this, intelligence cells have to process and evaluate information from all kinds of sources. Particularly in the area of non-conventional conflicts, e.g. in the fight against terrorism, heterogeneous and complex non-military information factors are influencing the production of intelligence. Large volumes of information and data have to be processed but manual evaluation and the conventional presentation of results is much too time consuming. A first essential aspect in the automated support to the exploitation and fusion of information and data is to deal with very different natures of data: numerical data, usually in the form of simple tables, more complex structured data such as relational databases, semi-structured messages, totally unstructured texts. A second challenging aspect in the automation of information fusion is provided by the heuristic nature of the real human processing of the imperfect information that is available. Fragments of information about ground truth are compared with the current picture of the situation, related to the most likely aspects of estimated threats, aggregated to more complex and significant situation objects, and finally integrated into a new picture of the situation. Correlation and aggregation of information is typically driven by knowledge about the structure and behaviour of adversary factions. This paper will present some of the findings of the RTO Task Group on Information Fusion Demonstration (IST-038/RTG-016) concerning these areas of topical interest. Relevant open research aspects and an analysis and functional model approach to human Intelligence processing is presented.