Multi-level Fusion of Hard and Soft Information for Intelligence

Driven by the underlying need for an as yet undeveloped framework for fusing heterogeneous data and information at different semantic levels coming from both sensory and human sources, we present some results of the research conducted within the NATO Research Task Group IST-106/RTG-051 on “Information Filtering and Multi Source Information Fusion.” As part of this ongoing effort, we discuss here a first outcome of our investigation on multi-level fusion. It deals with removing the first hurdle between data/information sources and processes being at different levels: representation. Our contention here is that a common representation and description framework is the premise for enabling processing overarching different semantic levels. To this end, we discuss here the use of the Battle Management Language (BML) as a way (“lingua franca”) to encode sensor- and text-based data and a priori and contextual knowledge, both as hard and soft data. We here expand on our previous works [1, 2] further detailing and exemplifying the use of BML and clarifying aspects related to the use of contextual information and the exploitation of uncertain soft input along with sensor readings.