Provided is a process including: sampling of nodes of a first graph; and forming a second graph comprising nodes, each of which corresponds to at least one entity mentioned in unstructured text documents. Forming the second graph comprises, for each pair of the nodes of the second graph, setting a relationship between the pair, the pair comprising (i) a first node corresponding to a first entity and (ii) a second node corresponding to a second entity, by: identifying, in the first graph, a node corresponding to an unstructured text document in which the first entity is mentioned; and setting the relationship between the pair based on a determination of whether or which of the sampled nodes comprises a node satisfying a set of criteria, the criteria comprising (i) having a relationship with the identified node and (ii) corresponding to an unstructured text document in which the second entity is mentioned.