Which is the best?: Re-ranking Answers Merged from Multiple Web Sources

The main motivation of this paper is to devise a way to select the best answers collected from multiple web sources. Depending on questions, we need to combine multiple QA modules. To this end, we analyze real-life questions for their characteristics and classify them into different domains and genres. In the proposed distributed QA framework, local optimal answers are selected by several specialized sub-QAs. For fining global optimal answers, merged candidates are re-ranked by adjusting confidence weights based on the question analysis. We adopt the idea of the margin separation of SVM classification algorithm to adjust confidence weights calculated by own ranking methods in sub-QAs. We also prove the effects of the proposed re-ranking algorithm based on a series of experiments.