Search Engine Result Aggregation Using Analytical Hierarchy Process

A metasearch engine queries search engines and collates information returned by them in one result set for the user. Metasearch can be external or internal. In external metasearch, result lists from external, independent search engines are merged. On the other hand, in an internal metasearch, result lists from using different search algorithms on the same corpus are aggregated. Thus result merging is a key function of metasearch. In this work, we propose a model for result merging that is based on the Analytic Hierarchy Process and compares documents and search engines in pair-wise comparison before merging. Our model has the capability to merge result lists based on ranks as well as scores, as returned by search engines. We use the LETOR 2 (LEarning TO Rank) dataset from Microsoft Research Asia for our experiments. When using document ranks, our model improves by 31.60% and 8.58% over the Borda-Fuse and Weighted Borda-Fuse models respectively. When using document scores the improvements are 42.92% and 18.03% respectively.