Filling the gaps in keyword-based query expansion for geodata retrieval

Abstract

Query expansion describes the automated process of supplementing a user's search with additional terms or geographic locations to make it more appropriate for the user's needs. Such process relies on the system's knowledge about the relation between geographic terms and places. Geodata repositories host spatial data, which can be queried over their metadata, such as keywords. One way to organize the system's knowledge structure for keyword-based query expansion is to use a similarity network. In a complete similarity network the total number of similarity values between keyterms increases with the square of included keywords. Thus, the task of determining all these values becomes time consuming very quickly. One efficient method is to start with a sparse similarity network, and automatically estimate missing similarity values from other values with an algorithm. Hence, this paper introduces and evaluates four such algorithms. © 2006 Springer-Verlag Berlin Heidelberg.