Mechanisms for managing an electronic document collection are provided. A first electronic document is analyzed to identify a reference to a second electronic document and the second electronic document is analyzed to identify document dependencies with zero or more other electronic documents. A dependency information data structure is generated based on the analysis. The dependency information data structure is analyzed to identify a subset of the electronic document collection that is to be loaded into memory when performing an information analysis operation. An electronic document curation action recommendation is generated based on the identified subset of the electronic document collection. The electronic document curation action recommendation is then output.