Fuzzy result merging models for metasearch.
详细信息
文摘
Each search engine queried by a metasearch engine returns results in the form of a result list of documents. The key issue is to combine these lists to achieve the best performance. In our work, we apply fuzzy aggregation operators to result merging. Our work is an extension of Yager's [45] fuzzy Ordered Weighted Average (OWA) operator based result merging model proposes by Diaz [13, 14]. We propose three extensions to the OWA model for metasearch. These are the Importance Guided OWA (IGOWA), the algebraic t-norm OWA, and the algebraic t-norm IGOWA models. While the first two are based on Yager's extension of the OWA operator, the third is a combination of the first two. The first model (IGOWA) allows weights to be applied to search engine result lists. The second model (t-norm OWA) allows for alternative t-norm functions to be used in aggregation. The third t-norm IGOWA model allows for both. In our work, for the second and third models we use the algebraic (product) t-norm. We compare and contrast our models and also compare them with existing models such as the OWA model for metasearch proposed by Diaz [13, 14] and the Borda-Fuse model proposed by Aslam and Montague [2].;Two of our models, the algebraic t-norm IGOWA model and the IGOWA model, require search engine weights. Thus we develop a new scheme for obtaining search engine weights. We apply our scheme to the above models and observe that using our weighting scheme results in improved result merging.;Lastly, Diaz [13, 14] provides heuristics for handling missing documents in individual search engine result lists before result merging. We apply the same concepts of the OWA and IGOWA to develop fuzzy heuristics for handling missing documents. Our experimental analysis illustrates under what conditions the proposed fuzzy heuristics for handling missing documents leads to enhanced performance of the result merging models.