Systems and methods are provided for processor-assisted indexing and searching, in which categories or individual documents within a database are annotated to the terms of an ontology, searched and then prioritized by rank. Items of the domain of interest, as well as the ontologies that describe attributes of those items, are embedded into a Bayesian network. Search queries correspond to user activations of items. An error model mimics incomplete or falsely specific items. Search results correspond to a subset of the items to which the query items are semantically-related. A probabilistic inference is performed to obtain marginal probabilities of each search result to explain the presence of the queries. These probabilities can then be used for ranking the search results.