Conceptual query expansion

Without detailed knowledge of a collection, most users find it difficult to formulate effective queries. In fact, as observed from Web search engines, users may spend large amounts of time reformulating their queries in order to satisfy their information need. A proven successful method to overcome this difficulty is to treat the query as an initial attempt to retrieve information and use it to construct a new, hopefully better query. Another way to improve a query is to use global (thesauri-like) information. In this article a new, hybrid approach is presented that projects the initial query result on the global information, leading to a local conceptual overview. The resulting concepts are candidates for query refinement. To show its effectiveness, we show that the conceptual structure resulting after a typical short query (2 terms) contains refinements that perform as well as a most accurate query formulation. Next we show that query by navigation is an effective mechanism that in most cases finds the optimal concept in a small number of steps. If the optimal concept is not found, then it still finds an acceptable sub-optimum. We show that the proposed method compares favorably to existing techniques.