AP-Scored Borda Counting for Information Retrieval

Borda counting is a rank aggregation method widely used in the information retrieval area. Despite many variants of Borda counting, the problem of how to take into account the prior structural information of rank lists is still open. In this paper, we analyze the structure of a list given its AP (average precision), and then propose the AP-scored Borda counting method under the information retrieval scenario. By properly defining the elementary score function, the structural information of each list, which is inferred from its AP, is naturally incorporated into the Borda counting procedure. Rank aggregation experiments on both simulated data and TRECVID 2004 data show promising results