Next: Background Up: Introduction Previous: Introduction Contents
Proposals for the union between logic and the probabilities theory go back to the research of George Boole in the last century [1], and have been discussed frequently in the last decades [2,3]. The reason for this interest is the potential generality of this unifying language, with potential applications in knowledge representation, objects description in group technology, expert systems for diagnostics, learning from rules, information search and recovery, description of resources distributed in a network [4] and planning under uncertainty.
Nowadays there as two main approaches for the combination of logic and probabilities. The first one basically associates probabilities to general sentences expressed in some logic. The second (and more recent) one starts by restricting the language and assuming independence relations so as to guarantee uniqueness in probabilistic assessments. These two approaches are, in many aspects, complementary. Probabilistic logic offers big flexibility, but typically dispense the representation of independence relations; on the other hand, the ``relational probabilistic model'' uses independence relations, but offers limited flexibility.