A query-aware document ranking method for geographic information retrieval

Geographically oriented search must consider both the thematic and geographic dimensions of relevance when matching documents to queries. We propose a dynamic document ranking scheme to combine the thematic and geographic relevance measures on a per-query basis. Query specificity is introduced to determine the relative weights of different sources of ranking evidence for each query. A preliminary evaluation comparing with human judgment shows that our method to distinguish different types of geo-referenced queries based on query specificity is promising to address the issue of relevance combination in GIR document ranking. In addition, we explore the possibility of using Dempster-Shafer's theory to combine the two different sources of ranking evidence.