Abstract

The need to meaningfully combine sets of rankings often comes up when one deals with ranked data. Although a number of heuristic and supervised learning approaches to rank aggregation exist, they generally require either domain knowledge or supervised ranked data, both of which are expensive to acquire. To address these limitations, we propose 1 a mathematical and algorithmic framework for learning to aggregate (partial) rankings in an unsupervised setting, and instantiate it for the cases of combining permutations and combining top-k lists. Furthermore, we also derive an unsupervised learning algorithm for rank aggregation (ULARA), which approximates the behavior of this framework by directly optimizing the weighted Borda count. We experimentally demonstrate the effectiveness of both approaches on the data fusion task.

Citations