In this paper, a methodology and generic algorithms are proposed to discover and to qualify authority links (a kind of coreference links) between entities, mentioned in descriptions of documents, and authorities, i.e., entities described in referential bases. This framework has been instantiated on the French national database Sudoc built and maintained by ABES (the French National Agency for Academic Libraries). Experiments demonstrate the effective usefulness of the tool developed using this methodology and ABES is finalizing an interactive tool for professional librarians to qualify existing links and also, if needed, to correct and complete these links.

We study the expressiveness of Nested Graphs, an extension of conceptual graphs. Nesting is introduced as a formal version of the intuitive “zooming in” on descriptions of individuals. Projections are defined inductively as the formal tool for “reasoning with nested graphs.” Nested graphs are translated to “colored” formulas. Coloring represents anaphoras in a way similar to conceptual graphs. A system of Gentzen sequents is shown to be adequate and complete with respect to projections of nested graphs.