A Bayesian network is a probabilistic model that is used to specify a set of events
and the dependencies between them. The networks are directed, acyclic graphs
(DAGs), where the nodes in the graph represent events with a set of possible
outcomes and arcs represent probabilistic dependencies between the events. The
probability, or belief,15 of a particular event outcome can be determined given
the probabilities of the parent events (or a prior probability in the case of a root
node). When used as a retrieval model, the nodes represent events such as observing
a particular document, or a particular piece of evidence, or some combination
of pieces of evidence. These events are all binary, meaning that TRUE and FALSE
are the only possible outcomes.