Application of Data Fusion in the Web Track.
Chunlan Huang, Shengli Wu, Jinbo Feng, Yongquan Tao, Yuping Xing · Text REtrieval Conference · 2013
In this paper, we report on the experiments we conducted whilst participating in the TREC 2013 Web track. We use data fusion to test how to improve the results from different information retrieval systems. Linear combination is used for fusion and multiple linear regression is used to obtain suitable weights for all the component systems involved. In our experiments, the ClueWeb09 dataset is used as training data to obtain weights for the three component systems Indri, MG4J, and Terrier. After running the official evaluation program, we find that all four runs submitted are better than all component results with one exception.