Using relevance to train a linear mixture of experts
作者： Christopher C Vogt , Garrison W Cottrell , Richard K Belew , Brian T Bartell , None
DOI:
关键词: Linear combination 、 Data mining 、 Statistic 、 Computer science 、 Vector space 、 Relevance (information retrieval) 、 Mixture of experts
摘要: A linear mixture of experts is used to combine three standard IR systems. The parameters for the are determined automatically through training on document relevance assessments via optimization a rank-order statistic which empirically correlated with average precision. improves performance in some cases and degrades it others, degradations possibly due techniques, model strength, poor individual
参考文章(0)