Effect of combining different Web search engines on Web question-answering

We have several different commercial Web search engines that are available. It is ex-pected that the combination of the results from different Web search engines has some impact on the accuracy of question-answering (QA) for Web documents, be-cause the search results are not identical and the combination increases the variety of information source. However, as far as we know, there are no studies on the ef-fect of combining different Web search en-gines on QA. In this paper, we examined the effect of the combination on QA. We investi-gated three different methods to combine search results from different search en-gines in the process of QA. The first one is a conventional method that straightfor-wardly merges search results from differ-ent search engines, then, feeds the unified search result into one QA engine. On the other hand, the second one and third one are our proposal methods that feed each search result from individual search en-gine to a QA engine separately, then merge the answer candidates. The experimental result showed that the methods that merge the answer candidates after QA are more effective than the method that merges the search results before QA.