Most enterprise systems rely on relational databases, and therefore SQL queries, to populate dynamic documents such as business intelligence reports, dashboards or spreadsheets. These queries represent metadata about the documents, thus they can feed information retrieval systems such as recommender systems or search engines. In this paper we propose to automatically annotate documents with structured representations of their SQL queries expressed with RDF graphs. We show that SPARQL is a natural language to query these SQL queries, i.e. to perform meta-querying, and discuss challenges that arise from this approach.