Data representation and matching for events and templates
J.R. Gabriel, Mohamed Taha Gabriel · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1988
This note examines means and reasons to encode event descriptions into bit strings - or, rather, strings of decision flags taking the values TRUE, FALSE, or UNKNOWN. In the case of perfect knowledge, all flags are TRUE or FALSE. In variants of the theory where degrees of confidence are taken into account, UNKNOWN is replaced by a value in the interval (0:1) representing the degree of confidence that the TRUE branch of the decision matches ground truth or the result of perfect knowledge. In templates, a fourth value - ARB - may be used, indicating an event attribute the same as or related to attributes of other events in the same template. We consider here the encoding of events whose place, time, and other attributes are known. Specifically we show that any set of events so encoded can be sorted to facilitate search for patterns such as troop concentrations. Templates are also sets of events with some attributes UNKNOWN and others ARB in the sense that their values are determined once the values of ARB's for other events in the template are known from matching (unification) with observed events in a history. Thus templates containing an ARB are properly speakingmore » pattern matching algorithms, and can in principle be complex procedures. A match with an UNKNOWN attribute in a template can be TRUE or FALSE and, unlike ARB, does not propagate a value into other attributes. The possibility of linearly ordering the events in a history and perhaps in a template also enables one to search an event file for aggregates to be done by passing ''windows'' over the event file.« less