Existing tools for library search typically support queries that directly exploit document properties such as title, author and subject. A lot of work in Knowledge Organization (KO) and Knowledge Representation (KR) shows that there is a need for more expressive queries where it is possible to predicate about the properties of relevant entities and, in particular, of those which are mentioned in the subject. These entities include people, organizations, locations, and events. Our proposed solution consists in the definition and exploitation of an entity model for FRBR, called eFRBR, which captures the main entities and properties informally defined in FRBR. When a specific use case needs to be addressed, eFRBR allows for the construction of a knowledge graph which codifies all the relevant entities which can then be suitably queried to obtain the desired results. We prove the validity of our approach in a use case addressing the scientific production at the University of Trento in Italy.