This research explores the feasibility of performing passive information capture on voice data in order to analyze and classify the contents of interpersonal communication. The general form of this problem is very difficult as fully automated speech understanding technology does not exist. This is further complicated by battlefield realities including: noise, jargon and unstructured speech. However, when specific topics are isolated for extraction, the challenge becomes manageable. Conceptual Spaces is used as a fusion framework to classify data passively captured by traditional speech recognition software coupled with fuzzy logic to provide matching of phonetics to jargon. Together these technologies prove to be a valuable fusion framework because of their ability to mitigate the high levels of errors inherent in speech recognition. An initial study focused on recognizing important topics in communications between commanders and field personnel amidst background chatter. Results indicate the Conceptual Spaces model is flexible enough to define “spaces” for military events, and the underlying optimization model used for classification was robust and fast enough to quickly and accurately classify the noisy scenario data. This technology enables a new and more general class of automation, permitting conversion of passive speech into structured data.