Numerous historical works are devoted to the life and works of Henri Poincaré (1854-1912), in particular through the study and the publication of his correspondence, which consists of about 2000 letters and includes academic, private and scientific exchanges. For several years, digital projects have been carried out to store, publish and exploit corpus data by implementing standards and technologies of the Semantic Web, including RDF, RDFS and SPARQL. When browsing an RDF graph, several situations may lead to a desire of flexible querying. This term describes search methods that go beyond conventional search systems, which are restricted to exact queries and allow limited or no expression of user preferences. The main contribution of this research work is the formalization, study and applications of a flexible query mechanism based on the use of SPARQL query transformation rules. This system allows, from an initial query, an RDF graph and a set of rules, to generate SPARQL queries which can provide alternative results to those initially returned. Some rules are generic, and therefore easily transposable to other graphs, and other rules are domain-dependent. Several tools based on this mechanism have been developed to assist the digital exploitation of the Henri Poincaré correspondence. A system has been implemented to assist the manual editing of RDF data, a task which can sometimes be tedious. This system relies on domain knowledge and the use of case-based reasoning to provide an ordered list of suggestions when editing an RDF triple. The proposed flexible querying system has also been integrated into a navigation tool, which provides an interface for visual exploration of RDF graphs, and which exploits similarities between resources in a graph to generate search filters. These tools exploit knowledge associated with the correspondence corpus which is represented through various transformation rules. Through the use of this mechanism, this work also considers the evolution of research practices in history, and tends to show how such a flexible querying system can contribute to the heuristic method. The methods and tools proposed can be applied to other corpora, in particular in the context of digital humanities projects.