Probabilistic Relevance Models Based on Document and Query Generation
Probabilistic Relevance Models Based on Document and Query Generation
复制标题
DOI：
10.1007/978-94-017-0171-6_1
复制
发表时间：
2003
期刊：
影响因子：
--
通讯作者：
J. Lafferty;Chengxiang Zhai
中科院分区：
文献类型：
--
作者：
J. Lafferty;Chengxiang Zhai
我们给出了一个统一的帐户的概率语义的语言建模方法和传统的概率模型的信息检索，这两种方法可以被视为是等价的概率，因为它们是基于不同的因子分解相同的生成相关模型。我们还讨论了这两种方法如何导致不同的检索框架在实践中，因为它们涉及组件模型，估计完全不同。
We give a unified account of the probabilistic semantics underlying the language modeling approach and the traditional probabilistic model for information retrieval, showing that the two approaches can be viewed as being equivalent probabilistically, since they are based on different factorizations of the same generative relevance model. We also discuss how the two approaches lead to different retrieval frameworks in practice, since they involve component models that are estimated quite differently.