Study on the Combination of Probabilistic and Boolean IR Models for WWW Documents Retrieval
暂无分享,去创建一个
In this paper, we describe our information retrieval (IR) system that is used for the NTCIR-4 Web Task A. First, we introduce our IR system, which is based on the probabilistic IR model. This system is quite similar to the Okapi system, and uses both a word index and a phrase index comprising combinations of two adjacent words. Second, we propose a method for clarifying queries that combines the probabilistic IR model and the Boolean IR model. Since it is not easy to construct a Boolean query that covers all relevant documents, a mechanism for clarifying the Boolean query is required. In this paper, we propose “appropriate Boolean query reformulation for IR” (ABRIR) that support Boolean query formation and score documents based on combining probabilistic and Boolean IR models. Finally, we discuss the effectiveness of the method based on the results of experiments.
[1] Keizo Oyama,et al. Overview of the Informational Retrieval Task at NTCIR-4 WEB , 2004, NTCIR.
[2] Masaru Kitsuregawa,et al. University of Tokyo/RICOH at NTCIR-3 Web Retrieval Task , 2002, NTCIR.
[3] Stephen E. Robertson,et al. Okapi/Keenbow at TREC-8 , 1999, TREC.