数字图书馆技术的渊源进展和反思材料.pptVIP

* Spearman rank correlation coefficient is the the method to Comparing two rankings on the same set of items in statistics . If the agreement between two rankings is perfect, the value is 1. If the one ranking is completely diverse of the other, the value is -1. The increasing values between them imply increasing agreement between the rankings. Problem Document/Query Geographic Scope Thematic Scope Footprints (coordinates) Spatial indexing Spatial relevance Keywords (term vectors) Textual indexing Thematic relevance Geographic Representation Thematic Representation Problem Document/Query Geographic Scope Thematic Scope Footprints (coordinates) Spatial indexing Spatial relevance Keywords (term vectors) Textual indexing Vector space relevance Geographic Representation Thematic Representation One relevance score How? Related Work Current methods: (Martins et al. 2005)(Cai, 2002)(Vogt Cottrell, 1999) Weighted sum of individual relevance scores Product of individual scores Maximum (winner take all) Step-linear function A static formula is used to combine geo- and thematic relevance scores. -- A problem! Andrade and Silva (2006) showed that the relative importance of thematic and geographic relevance is query-dependent Our Work Infer the relative importance of geographic and thematic relevance by analyzing the nature of the query. Users tend to use more specific terms when they emphasize an aspect of the query. That leads to the notion of query specificity. Method The notion of query specificity: how specific (or general) a query is. Thematic specificity: how specific the thematic terms in a query are Geographic specificity: how specific the geographic scope in a query is Assumption: Thematic or geographic specificity reveals the relative power of corresponding relevance score in determining the document ranking. Method Query Thematic specificity Geographic specificity Q1: “Hilton hotels in central Pennsylvania” High Low Q2: “Hotels along the A