Learning how to balance semantic density and semantic overlap is essential for machine retrieval and for readers