Situation assessment is the name given to the activity of evaluating a given sequence of events to make a coherent and abstracted interpretation and predicting coming events based on the interpretation. A situation can be modelled by a script-a sequence of stereotypical events. In that case, situation assessment task can be formulated as script matching in that reported events are matched to model scripts to check whether the situation of the model script has occurred. Events are treated as a clue for invoking hypotheses as well as evidence for confirming hypotheses. A hypothesis may be constructed with lower level hypotheses.
In this thesis, a script structure is defined and a generic script matching methodology is proposed. Based on the methodology, a script-based situation assessment tool called SMAT (Script Matcher) is developed. SMAT creates hypotheses with reported events, generates expectations of the invoked hypotheses and confirms the hypotheses by matching coming events with the expectations.
The script structure is capable to represent various situations by providing the facilities of defining local variables in a script and sharing the lower level scripts with other scripts. The likelihood of a script is defined formally as a weighted sum of matching probability of script components and reported events. The matching process, which is applied recursively to hierachically structured hypotheses, has been shown suitable to manage complex hypotheses of multiple levels. In conclusion, SMAT provides a useful tool for construction of script-based situation assessment expert systems.
