We explore how Virtual Research Environments based on Semantic Web technologies support research interactions with RDF data in various stages of corpus-based analysis, analyze the Web of Data in terms of human readability, derive labels from variables in SPARQL queries, apply Natural Language Generation to improve user interfaces to the Web of Data by verbalizing SPARQL queries and RDF graphs, and present a method to automatically induce RDF graph verbalization templates via distant supervision.