Another method of computing the similarity function r is probabilisticly. There are two popular probabilistic models that we will discuss: the traditional Robertson / Sparck Jones formalization and the more structural inference network model.
