Query reformulation with collaborative concept-based expansion

Most information retrieval systems on the Internet try to supply the ‘best’ documents to a certain query by ranking the complete hit list with various algorithms. But this is only appropriate if the hit list contain relevant documents which can be positioned on the top. In the case of a small or empty hit list no ranking can improve the result and the user must reformulate the query again and again until he gets a sufficient result. This is a tedious task especially if the query is imprecise or too specific. We propose a method for improving the original query by an automatic reformulation method. Each phrase of the query corresponds to a concept where similar terms are stored. These terms are used to reformulate the original query and the user directly receives a hit list with more relevant documents without numerous searching circles.