This article describes the development of methods for identifying semantically significant documents in sets of texts of court decisions in the context of information retrieval. We considered the basic principles of construction of the graph of citations on the basis of legal text sets. We also described the applied network metrics for the analysis of the constructed citation graph. On the constructed model basis, an experiment was conducted in which the results of the semantic analysis of the model were evaluated by experts. As a result, we presented a software package that allows to search for semantically significant documents in court decisions text sets.