שליחת רשומה: Efficient Queries of Stand-off Annotations for Natural Language Processing on Electronic Medical Records