基于Dempster-Shafer理论的查询扩展模型

查询扩展是提高信息检索召回率的有效方法，已有许多研究者提出了诸多扩展方法，但大多数方法均是简单地把扩展的词添加到查询中．这样如若不加区分原查询词和扩充词，则扩展后的查询就可能偏离原查询的语义．该文基于Dempster-Shafer证据理论，提出了一个查询扩展的计算模型，该模型把原始查询项看作主证据，扩展项为原始查询项的辅助证据，为了能组合这两类证据，该文给出了两个带权重的Dempster-Shafer组合规则，实验表明该方法能有效地提高检索效率．

A Dempster- Shafer Model for Query Expansion

Query expansion is an effective way to extend the coverage of retrieval to the related documents. Various approaches have been proposed and many of them are simply adding expansion terms into the original query terms. In our early work we have showed that these simplistic expansion method can bias the focus of the original query, because the expanded terms add additional emphasis to the original term. Instead of adding expansion terms into the vector, we propose to combine them with the original terms by means of a logical OR operator. In this paper, we propose a query expansion computing model within the Dempster-Shafer Theory of Evidence.In the model, the original query terms are considered as primary evidences and the expansion terms are considered as auxiliary evidences to the original terms, and improve the Dempster-Shafer rule for query expansion.Experiments using the TREC collection show that our expansion method is more appropriate than the simple addition approach.