Abstract

Document search is often implemented by assigning a score to each document and then searching the documents with the highest scores. Sometimes, constraints are added to the search, such as searching documents with a maximal average age, or needing the resulting documents to be as diverse a possible. Several such constraints are formulated as a mixed integer linear program. The resulting problem is then solved by using wellknown algorithms from linear programming. Various possible constraints are researched, combined, and then compared to existing algorithms and implementations.