A document-centric approach to static index pruning in text retrieval systems