SPIRIT and Léa Sombé: A study in probabilistic reasoning

In this paper, the famous Léa Sombé-example is re-examined by methods combining cross-entropy minimization with probabilistic conditional logic. Thus a knowledge base is built up which allows easy computations and inferences in a logically sound way. The concept is realized by the probabilistic expert system shell SPIRIT which is presented here, too. So the aim of this paper affects as much practical aspects as it concerns logical foundations of knowledge representation. As the Sombé-example illustrates, even incomplete knowledge based on subjective probabilities or statistical data may be represented and dealt with adequately.