Model of Intention Inference Using Bayesian Network
Next we need to model human knowledge. It is known that an expert has well-designed knowledge composed hierarchically. Therefore we modelled such knowledge by using state hierarchy that is constructed from state nodes and links. Each state node represents an abstract concept and a link between two state nodes represents any relationship. In state hierarchy, the above state stands for a general concept and the below state a specific concept. In other words occurrence of the below state supports occurrence of the above. According to this relation we can get the graph that shows causality of state occurrence. In this graph each state has some symptoms that will be observed when the state occurs. In our model links between some child state and its parent represent OR relation, and those between a symptom and its parents noisy-OR relation. When S and X/ represent a state and a symptom respectively, noisy-OR relation is shown by the next equations.