The inference network retrieval model, which has been used in both commercial and open source search engines (and is incorporated in Galago), is one approach to doing this.
The inference network model is based on the formalism of Bayesian networks
and is a probabilistic model. The model provides a mechanism for defining and
evaluating operators in a query language. Some of these operators are used to specify
types of evidence, and others describe how it should be combined. The version
of the inference network we will describe uses language models to estimate the
probabilities that are needed to evaluate the queries.
In this section, we first give an overview of the inference network model, and
then show how that model is used as the basis of a powerful query language for
search applications. In the next section, we describe web search and explain how
the inference network model would be used to combine the many sources of evidence required for effective ranking.