Relevance Measures Using Geographic Scopes and Types

This paper proposes two kinds of relevance measures to rank documents by geographic restriction: scope-based and type-based. The non-geographic and geographic relevance scores are combined using a weighted harmonic mean. The proposed relevance measures and weighting schemes are evaluated on GeoCLEF 2007 dataset with encouraging performance over the standard IR performance. The best performance is achieved when the importance of non-geographic relevance scores outweigh the importance of geographic relevance scores.