A decision mechanism for the selective combination of evidence in topic distillation

Abstract

The combination of evidence can increase retrieval effectiveness. In this paper, we investigate the effectiveness of a decision mechanism for the selective combination of evidence for Web Information Retrieval and particularly for topic distillation. We introduce two measures of a query's broadness and use them to select an appropriate combination of evidence for each query. The results from our experiments show that there is a statistically significant association between the output of the decision mechanism and the relative effectiveness of the different combinations of evidence. Moreover, we show that the proposed methodology can be applied in an operational setting, where relevance information is not available, by setting the decision mechanism's thresholds automatically.