Using First-Order Probability Logic for the Construction of Bayesian Networks

We present a mechanism for constructing graphical models speci cally Bayesian net works from a knowledge base of general probabilistic information The unique fea ture of our approach is that it uses a power ful rst order probabilistic logic for express ing the general knowledge base This logic allows for the representation of a wide range of logical and probabilistic information The model construction procedure we propose uses notions from direct inference to identify pieces of local statistical information from the knowledge base that are most appropri ate to the particular event we want to reason about These pieces are composed to gener ate a joint probability distribution speci ed as a Bayesian network Although there are fundamental di culties in dealing with fully general knowledge our procedure is practical for quite rich knowledge bases and it supports the construction of a far wider range of net works than allowed for by current template technology